1: %-----------------------This is LaTeX file----------------
2: \documentclass[a4paper,12pt]{article}
3: \usepackage[dvips]{graphicx}
4:
5:
6:
7: \font\Sets=msbm10
8: \def\Natural{\hbox{\Sets N}}
9: \def\Integer{\hbox{\Sets Z}}
10: \def\Real{\hbox{\Sets R}}
11: \def\Complex{\hbox{\Sets C}}
12: \def\bb{\begin{equation}}
13: \def\ee{\end{equation}}
14: \def\ban{\begin{eqnarray}}
15: \def\ean{\end{eqnarray}}
16: \def\ba{\begin{eqnarray*}}
17: \def\ea{\end{eqnarray*}}
18: \def\const{\hbox{const}}
19: \def\res{\hbox{res}}
20: \def\pt{\partial}
21: \def\dt{{d\over{dt}}}
22: \def\VPint{\backslash\!\!\!\!\!\int}
23: \def\a{\alpha}
24: \def\b{\beta}
25: \def\D{\Delta}
26: \def\d{\delta}
27: \def\G{\Gamma}
28: \def\g{\gamma}
29: \def\ve{\varepsilon}
30: \def\k{\kappa}
31: \def\l{\lambda}
32: \def\m{\mu}
33: \def\n{\nu}
34: \def\vf{\varphi}
35: \def\o{\omega}
36: \def\O{\Omega}
37: \def\r{\rho}
38: \def\s{\sigma}
39: \def\S{\Sigma}
40: \def\t{\tau}
41: \def\th{\theta}
42: \def\z{\zeta}
43: \newtheorem{demo}{Remark}
44:
45:
46: \title{\bf Asymptotic description of nonlinear
47: resonance \footnote{This work was supported by grants RFBR
48: (00-01-00663, 00-15-96038) and INTAS (99-1068)}}
49:
50: \author{O.M. Kiselev\footnote{Institute of Mathematics
51: of Ufa Sci. Centre of RAS;\ e-mail:ok@ufanet.ru} \and S.G.
52: Glebov\footnote{Ufa State Petroleum Technical University;\
53: e-mail:gloomy69@ufanet.ru}}
54: \date{May 2, 2001}
55:
56:
57: \begin{document}
58: \maketitle
59:
60: \begin{abstract}
61: We study a hard regime of stimulation of two-frequency
62: oscillations in the main resonance equation with a fast
63: oscillating external force: $ \ve i \psi' + |\psi|^2\psi =
64: \exp\big(it^2/ (2\ve)\big),\,\, 0<\ve\ll1$. This phenomenon is
65: caused by resonance between an eigenmode and the external
66: force. The asymptotic solution before, inside and after the
67: resonance layer is studied in detail and matched.
68: \end{abstract}
69:
70:
71:
72: \section{Introduction}
73: \par
74: In this paper we investigate the hard mode of stimulating of
75: two phase oscillations. We study this phenomenon in
76: asymptotic solution of ordinary nonlinear differential
77: equation under fast oscillating external force:
78: \bb
79: \ve i \psi' + |\psi|^2\psi = \exp\bigg({{it^2}\over
80: {2\ve}}\bigg), \label{sch}
81: \ee
82: where $0<\ve\ll1$ -- a small parameter.
83: \par
84: The solution of the equation (\ref{sch}) constructed in this
85: paper oscillates with the frequency of the perturbation when
86: $t>t_*$. An amplitude of oscillations changes slowly. This
87: system changing corresponds to the stimulated oscillations of
88: nonlinear equations. The frequency of the oscillations changes
89: with time. From the other hand, the eigen-frequency of the
90: nonlinear equation depends on the amplitude. At certain moment
91: $t=t_*$ the eigen-frequency of the oscillations becomes equal
92: to the external force frequency. A resonance takes place in
93: the system. This leads to hard loss of stability of the
94: eigenmode of the oscillations and the system gets out of the
95: resonance. The asymptotic solution of equation (\ref{sch})
96: becomes two-phase. One phase relies to the oscillations
97: stimulated by the perturbation force and another one
98: corresponds to the oscillations
99: stimulated by the transition over the resonance.
100: \par
101: To explain the nature of the solution bifurcation and to construct
102: the asymptotic solution it is more convenient to investigate
103: the equation for the amplitude of the oscillations
104: $U=\psi\exp\big(-it^2/(2\ve)\big)$
105: \begin{equation}
106: \ve i U' + |U|^2 U - tU = 1. \label{sh}
107: \end{equation}
108: \par
109: If we begin to study the amplitude $U$ and consider the equation
110: (\ref{sh}) then constructing of the asymptotic solution is equivalent
111: to investigation of the bifurcation of the slowly varied equilibrium of
112: the equation (\ref{sh}). The theory of the bifurcation usually
113: investigates the behavior of a solution depending on external
114: parameter which is not connected with a variable of a
115: differential equation \cite{Andr,Arn}.
116: \par
117: The bifurcation of the equilibrium with varying parameter is
118: well known in physics \cite{Z-S}. This bifurcation is usually
119: illustrated by loss of stability of the equilibrium of
120: mechanical system, which is described by the second order
121: differential equation with slowly varying coefficients. In
122: this case the hard loss of stability takes place. The numeric
123: evaluations give the picture:
124: \begin{figure}[h]
125: %\includegraphics[width=10.16cm,height=6.7cm]{parametric.bmp}%for DVI-file
126: \includegraphics[angle=0, bb=0 0 300 200]{parametric.ps}%for ps-file
127: %\includegraphics[width=10.16cm,height=6.7cm]{parametric.png}%for PDF-file
128: \caption{The bifurcation solution of the equation (\ref{sh}).}
129: \end{figure}
130: \par
131: A description of the internal asymptotic structure of the
132: solution in the case of the hard loss of the stability is
133: connected with a construction of the asymptotics for the
134: solution of the Painlev\'e-1 equation \cite{Hab1}. Besides the
135: solution has a complicated asymptotic structure in the
136: transition layer where the main term of the asymptotics is
137: defined by four various expansions of different types. There
138: are: a special solution of the Painlev\'e-1 equation, a
139: sequence of separatrix solutions of the homogeneous equation
140: with "frozen" coefficients, a sequence of solutions of the
141: Weierstrass equation with the parameter $g_2=0$ and a sequence
142: of solutions of the Weierstrass equation with the parameter
143: $g_2\not=0$.
144: \par
145: Two of the first internal layers were found and studied in
146: \cite{Hab1}. Later, in \cite{ok1}, \cite{Dim-Hab} the
147: asymptotic structure of the transition layer was studied in
148: detail. It was shown that actually it is necessary to study
149: the sequence of the separatrix solutions of the homogeneous
150: equation with "frozen" coefficients. The third internal layer
151: was found in \cite{Dim-Hab}.
152: \par
153: However, these three layers are not enough for a passage
154: through the bifurcation interval. There exists a layer which
155: is defined by a solution of the Weierstrass equation with
156: the parameter $g_2\not=0$. This layer is found and studied
157: in this work due to the matching of all asymptotic solutions
158: before inside and after the bifurcation layer. The another
159: new result of this paper is the calculating of internal
160: variables in a sequence of layers, which are connected with
161: the separatrix solutions of the homogeneous equation with
162: "frozen" coefficients. For the Painlev\'e-2 equation with a
163: small parameter at a derivative these calculations have been
164: made in the work of one of the authors of this paper in
165: \cite{ok2}.
166: \par
167: Among other works devoted to the investigation of the
168: asymptotics solutions for nonlinear differential equations
169: with variable coefficients it is necessary to note the
170: investigations of passage through a separatrix in \cite
171: {Timof} -\cite {Hab-Ho}, where there were no confluent of the
172: slowly varying equilibrium. In particular, in work
173: \cite{Hab-Ho} the passage through a separatrix of equation
174: (\ref{sh}) with constant coefficients and small dissipation
175: was studied. In the case considered in this work the solution
176: of the equation for the amplitude also passes through a
177: separatrix, but at more complicated confluent saddle-center
178: equilibrium.
179: \par
180: Here, the asymptotic solution for the equation (\ref{sch}) in
181: the domain where the amplitude varies slowly ($t>t_*$) is
182: constructed by perturbation theory method and in domain where
183: the amplitude fast oscillates --- by the Krylov-Bogolyubov
184: method \cite{K-B, Kuz}. Thus, we reproduce the elegant
185: formulas for the asymptotic solutions of the equation
186: (\ref{sch}) obtained in the work \cite{Bour-Hab2}. All
187: asymptotics are matched \cite{I}.
188: \par
189: The structure of this paper is following. In the second
190: section the problem is formulated. In the third one, the
191: obtained results are represented. The fourth section is
192: devoted to a construction of an algebraic asymptotic solution
193: for the equation (\ref{sh}). The fifth section includes an
194: investigation of a bifurcation layer and studying of the
195: sequences of the internal expansions. In the sixth section
196: the fast oscillating solution of the equation (\ref{sh}) is
197: constructed and its asymptotics at the approach to a
198: bifurcation point is written.
199:
200: \section{Studied problem and its justification}
201: \subsection{Justification}
202: \par
203: The general solution of the equation (\ref{sch}) generally
204: speaking is unknown. But we can construct a set of formal
205: asymptotic solutions on a small parameter $\ve$. The simplest
206: kind of the solutions is the solutions oscillating with the
207: frequency of the external force. To obtain this kind of
208: solutions it is necessary to proceed the amplitude equation
209: (\ref{sh}). Further following the natural suggestion about
210: boundedness of the derivative in the equation (\ref{sh}) we
211: can obtain a nonlinear algebraic equation for the main term of
212: asymptotics $\stackrel{0}{U}(t)$:
213: \bb
214: |\stackrel{0}{U}|^2\stackrel{0}{U} - t\stackrel{0}{U} = 1.
215: \label{alg}
216: \ee
217: The number of the roots of this algebraic equation depends on
218: a parameter $t$. The roots can be written explicitly. There
219: exist a value of the parameter $t$ equals to
220: $t_*=3(1/2)^{2/3}$ so that the equation (\ref{alg}) has three
221: real roots at $t>t_*$. At $t=t_*=3(1/2)^{2/3}$ there is one
222: simple root and one multiple root $U_*=-(1/2)^{1/3}$. At
223: $t<t_*$ the equation (\ref{alg}) has the alone root.
224: \par
225: Let us consider the domain $t>t_*$ where there are three
226: different roots of the equation (\ref{alg}). Denote them
227: $U_k$, and $U_3(t)<U_2(t)<U_1(t)$. These roots correspond to
228: slowly varying equilibriums of the equation (\ref{sh}). Two
229: of the equilibriums are stable centers. The formal asymptotic
230: solutions with the main terms $U_1(t),\, U_2(t)$ correspond
231: to these stable equilibriums. The third equilibrium is a
232: saddle. There exists a parameter value $t=t_*$ at which one
233: of the centers coalesces with the saddle. At that moment the
234: saddle-center bifurcation takes place. At $t<t_*$ there
235: exists only one slowly varying equilibrium.
236: \par
237: This bifurcation may by explained on the example of an
238: autonomous equation with a "frozen" coefficient $T$:
239: $$
240: iV'+(|V|^2-T)V=1.
241: $$
242: \begin{figure}[h]
243: %\includegraphics[width=17cm,height=5.36cm]{ph_port.bmp}%for DVI-file
244: \includegraphics[angle=0, bb=0 0 380 100]{ph_port.ps}%for PS-file
245: %\includegraphics[width=17cm,height=5.36cm]{ph_port.png}%for PDF-file
246: \caption{The phase portraits for the equations with the
247: "frozen" coefficient at $T<t_*$, $T=t_*$ and $T>t_*$.}
248: \end{figure}
249: On the figure 2 one can see the phase plane of this equation.
250: On the left picture one equilibrium exists. On the middle
251: picture one can find two equilibrium positions and at last on
252: the right picture one can find three equilibrium positions.
253: \par
254: Choosing $U_k$ as the main term of the asymptotics in the
255: domain $t>t_*$ one can construct three different formal
256: asymptotic solutions in the form:
257: \bb
258: U(t;\ve) = \sum_{n=0}^{\infty}\ve^n\stackrel{n}{U}(t). \label{alas}
259: \ee
260: Each of them corresponds to the one of the slowly varying
261: equilibrium. The formal solution with the main term $U_1(t)$
262: doesn't change the structure when the variable $t$ changes and
263: we do not investigate this case. Solution based on $U_3(t)$ is
264: unstable on the whole interval $t>t_*$ and it is also out of
265: our consideration. We will investigate the asymptotic
266: solution based on $U_2(t)$ and the behavior of this asymptotic
267: solution after the saddle-center bifurcation.
268:
269: \subsection{Statement of the problem}
270:
271: In this work we solve the problem on constructing of the
272: formal asymptotic solution of the equation (\ref{sch}) in the
273: interval $t\in[t_*-C,t_*+C]$ where $ C=\const>0$ uniform on
274: $\ve$ . We suppose that the solution in the domain $t>t_*$ has
275: the form $$ \psi(t,\ve)=\exp\bigg({it^2\over2\ve}\bigg)
276: \sum_{n=0}^{\infty}\ve^n\stackrel{n}{U}(t),\quad \ve \to 0, $$
277: where $\stackrel{0}{U}(t)=U_2(t)$.
278:
279:
280: \section{Result}
281: \par
282: The combined asymptotic solution of the equation (\ref{sch})
283: in an interval $t\in[t_*-C,t_*+C]$ is constructed. The
284: solution has the different asymptotic structure in different
285: parts of the interval. The main result of this work is
286: following: firstly, the domain where constructed combined
287: solution is valid completely covers the interval
288: $[t_*-C,t_*+C]$ and secondly, all asymptotics are matched. In
289: this section we only represent the form of asymptotics and
290: leader terms. The explicit formulas can be found in the
291: corresponding sections of the paper.
292: \par
293: In the domain $(t-t_*)\ve^{-4/5} \gg 1$ the asymptotic solution
294: has the form: $$ \psi(t,\ve)= \exp\bigg({it^2\over2\ve}\bigg)
295: \sum_{n=0}^\infty\ve^{n}\stackrel{n}{U}(t),\quad \ve \to 0. $$
296: The leader term of the asymptotics is equal to the middle root of
297: algebraic equation (\ref{alg}): $\stackrel{0}{U}(t)=U_2(t)$. The
298: correction terms $\stackrel{n}{U}(t)$ are algebraic functions of
299: $t$.
300: \par
301: In the domain $|t-t_*|\ll 1$ the asymptotic solution is defined
302: by four various expansions of different types. First of them is:
303: \bb \psi=\bigg(U_* +
304: \ve^{2/5}\sum_{n=0}^\infty\ve^{2n/5}\stackrel{n}{\alpha}(\tau) +
305: i\ve^{3/5}\sum_{n=0}^\infty\ve^{2n/5}\stackrel{n}{\beta}(\tau)
306: \bigg)\exp\bigg({it^2\over2\ve}\bigg),\quad \ve \to 0, \label{as2}
307: \ee where $U_*$ is multiple root of the cubic equation
308: (\ref{alg}) at $t=t_*$. The coefficients of this expansion
309: depend on new scaling time $\t=(t-t_*)\ve^{-4/5}$. The leader
310: term of the asymptotics and the corrections are defined by
311: their asymptotics as $\t\to-\infty$ uniquely. In particular,
312: the leader term of asymptotics $\stackrel{0}{\a}(\t)$ is a
313: special solution of the Painlev\'e-1 equation: $$
314: \stackrel{0}{\alpha}{}^{\prime\prime} -
315: 3\stackrel{0}{\alpha}\!\!^2+\tau = 0. $$ with the given
316: asymptotics as $\t\to-\infty$: $$ \stackrel{0}{\alpha}(\tau) =
317: \sum_{n\ge0} \alpha_n \tau^{-{{(5n-1)}\over
318: 2}},\quad{\hbox{where}}\,\,\,\,
319: \a_0={1\over\sqrt{3}},\,\,\a_1={1\over24}. $$ In the domain
320: $\t>-\infty$ this solution has poles on the real axis of $\t$.
321: Denote the least of them by $\t_0$. The asymptotics
322: (\ref{as2}) is valid as $(\tau-\tau_0)\ve^{-1/5}\gg 1$.
323: \par
324: In the neighborhood of the bifurcation point $(\t=\t_0)$ the
325: coefficients of the asymptotic expansion depend on one more
326: fast time scale $\theta=(\t-\t_0)\ve^{-1/5}$. Denote by $$
327: \th_0=\th+\sum_{n=1}^\infty\ve^{n/5}\stackrel{n}{\th}_0, $$
328: where $\stackrel{n}{\th}_0=\const$. Then in the domain
329: $-\ve^{-1/5}\ll\th_0\ll\ve^{-1/10}$ the formal asymptotic
330: solution has the form: $$ \psi(t,\ve)
331: =\bigg(U_*+\stackrel{0}{w}(\th_0)+\ve^{4/5}
332: \sum_{n=1}^\infty\ve^{(n-1)/5}\stackrel{n}{w}(\th_0)\bigg)
333: \exp\bigg({it^2\over2\ve}\bigg), \quad \ve \to 0. $$ The main
334: term of asymptotics $\stackrel{0}{w}(\th_0)$ is the separatrix
335: solution of the autonomous equation: \bb
336: i\stackrel{0}{w}{}^\prime +
337: U_*\left(2|\stackrel{0}{w}|^2+\stackrel{0}{w}\!{}^2\right) +
338: U_*^2 \left( \stackrel{0}{w}\!^* - \stackrel{0}{w} \right) +
339: |\stackrel{0}{w}|^2 \stackrel{0}{w}= 0, \label{avt} \ee
340: namely: $$ \stackrel{0}{w}(\th_0)
341: ={{-2}\over{(\th_0-iU_*)^2}}. $$
342: \par
343: In the domain $-\th_0\gg1$ the asymptotic solution is defined
344: by a sequence of two alternating asymptotics. Let us call
345: them by "intermediate" and "separatrix" asymptotics. To obtain
346: the intermediate asymptotics let us introduce one more slow
347: variable:
348: $$ T_k=\th_{k-1}\ve^{1/6}, k=1,2,\dots. $$ An asymptotic solution
349: in the intermediate domain for not too large values $ k $: $$
350: k\ll\ve^{-1/7} $$ has the form:
351: $$
352: \psi(t,\ve)=\bigg(U_*+\ve^{1/3}\sum_{n=0}^{\infty}\ve^{1/30}
353: \stackrel{n}{A}_k(T_k)
354: \,+\,
355: $$
356: $$
357: i\ve^{1/2}\sum_{n=0}^{\infty}\ve^{1/30}
358: \stackrel{n}{B}_k(T_k)\bigg)\exp\bigg({it^2\over2\ve}\bigg),\quad
359: \ve \to 0. $$ The leader term of the asymptotics satisfies to
360: the equation: $$
361: \stackrel{0}{A}_k\!\!\!''+3\stackrel{0}{A}_k\!\!\!^2=0 $$ and
362: can be expressed by the Weierstrass function with the
363: parameter $g_2=0$:
364: \bb
365: \stackrel{0}{A}_k=-2\wp(T_k;0,g_3(k)),\,\,\,\hbox{where}
366: \,\,\, g_3(k)={1\over56}\big(g_3(k-1)+\pi/2\big),\,\,\,
367: g_3(0)={a_4\over56}. \label{A-0-0}
368: \ee
369: The constant $a_4$ is
370: the coefficient as $(\t-\t_0)^4$ in the Laurent expansion of
371: $\stackrel{0}{\a}(\t)$.
372: \par
373: The Weierstrass $\wp$-function has a real period $\O_k$ and
374: has poles on the real axis at $T_k=0$ and $T_k=\O_k$. The
375: intermediate expansion with the leader term (\ref{A-0-0}) is
376: valid in the domain between the poles as
377: $$
378: -\ve^{-1/6}T_k\gg1,\quad \ve^{-2/15}(T_k+\O_k)\gg1.
379: $$
380: \par
381: At the large values of $k$ the intermediate asymptotics are
382: constructed in the form
383: $$
384: \psi(t,\ve)=\bigg(U_*+\ve^{1/3}\sum_{n=0}^{\infty}\ve^{n/6}
385: \stackrel{5n}{A}_k(T_k)
386: \,+\,
387: $$
388: $$
389: i\ve^{1/2}\sum_{n=0}^{\infty}\ve^{n/6}
390: \stackrel{5n}{B}_k(T_k)\bigg)\exp\bigg({it^2\over2\ve}\bigg),\quad
391: \ve \to 0.
392: $$
393: The leader term of asymptotics satisfies to the
394: equation: $$
395: \stackrel{0}{A}_k''+3\stackrel{0}{A}_k\!\!\!^2=\l_k.
396: $$ $$ \l_k(\ve)=\ve^{1/6}\bigg(\sum_{j=1}^{k}\O_j
397: \,+\,\sum_{n=1}^\infty\ve^{(n-1)/30}
398: \sum_{j=1}^k\stackrel{n}{x}_j\!\!\!^+\bigg). $$ The main term
399: of the asyptotics is: \bb
400: \stackrel{0}{A}_k(T_k)=-2\wp(T_k,\l_k/2,g_3(k,\ve)),
401: \label{A-0-lk} \ee where $$
402: g_3(k,\ve)=\stackrel{0}{g}_3(k)+\sum_{n=1}^\infty
403: \ve^{n/30}\stackrel{n}{g}_3(k). $$ The intermediate expansion
404: with the leader term (\ref{A-0-lk}) is valid in the domain
405: between the poles of the Weierstrass function as $$
406: -\ve^{-1/6}T_k\gg1,\quad \ve^{-2/15}(T_k+\O_k)\gg1. $$
407: \par
408: The separatrix expansions are valid in a small neighborhood
409: of the Weierstrass function poles. Denote: $$
410: \th_k=\big(T_k+\O_k-{1\over4}\sum_{n=1}^\infty\ve^{n/30}
411: \stackrel{n}{x}_k\!\!\!^+\big)\ve^{-1/6},\quad k=1,2,\dots. $$
412: When $$ |\th_k|\ve^{1/6}\ll1 $$ the formal asymptotic
413: solution of equation (\ref{sch}) has the form: $$
414: \psi=\bigg(U_*+\stackrel{0}{W}(\th_k)+\ve^{4/5}\sum_{n=1}^\infty
415: \ve^{(n-1)/30}\stackrel{n}{W}(\th_k)\bigg)\exp\bigg({it^2\over2\ve}\bigg)\quad
416: \ve \to 0. $$ The leader term of the asymptotics
417: $\stackrel{0}{W}(\th_k)$ is a separatrix solution of the
418: autonomous equation (\ref{avt}): $$ \stackrel{0}{W}(\th_k)
419: ={{-2}\over{(\th_k-iU_*)^2}}. $$
420: \par
421: The sequence of the alternating intermediate expansions and
422: separatrix asymptotic ones is valid as
423: $\ve^{-1/6}(t_*-t)\ll1$.
424: \par
425: In the domain $(t_*-t)\ve^{-2/3}\gg1$ the asymptotic solution
426: becomes two-phase. The amplitude of the stimulated
427: oscillations in the solution of (\ref{sch}) oscillates fast.
428: The form of the solution is:
429: $$
430: \psi=\bigg(\stackrel{0}{U}(t_1,t,\ve) + \ve
431: \stackrel{1}{U}(t_1,t,\ve) + \ve^2 \stackrel{2}{U}(t_1,t,\ve)
432: \bigg)\exp\bigg({it^2\over\ve}\bigg),
433: $$
434: where $t_1$ is a new fast variable $t_1=S(t)/\ve+\phi(t)$.
435: The main term of the asymptotics $\stackrel{0}{U}$ lies on the
436: curve $\G(t)$:
437: $$
438: {1\over 2}|y|^4 -t |y|^2 - (y + \bar y) = E(t),
439: $$
440: and satisfies to the Cauchy problem for the equation
441: $$
442: iS^{\prime}\pt_{t_1} \stackrel{0}{U} = \pm
443: \sqrt{2\stackrel{0}{U}\!^3 + (2E(t)+t^2) \stackrel{0}{U}\!^2 +
444: 2t\stackrel{0}{U}+1},
445: $$
446: with an initial condition $\stackrel{0}{U}|_{t_1=0}=u_0$, such,
447: that $\Im(u_0)=0,\quad \Re(u_0)=\min_{y\in
448: \G(t)}\big(\Re(y)\big)$. The function $S(t)$ is a solution
449: for the Cauchy problem
450: $$
451: iS'\int_{\G(t)}{dy\over\sqrt{3y^3+(2E+t^2)y^2+2ty+1}}=T, S|_{t=0}=0,
452: $$
453: where $T=\const>0$. The function $E(t)$ is the solution of
454: the transcendental equation
455: $$
456: i\int_{\G(t)}u^*du=\pi.
457: $$
458: The phase shift $\phi$ is defined by two constant. One of
459: them is an "initial" condition $\phi(t_*)=\phi_0$ and another one
460: is a constant $\phi_1$ in the equation in for $\phi$:
461: $$
462: {\phi'\over \pt_E S}\pt_E I =\phi_1.
463: $$
464: \par
465: \begin{demo} In our work constants $\phi_0$ and $\phi_1$ are not
466: calculated. It means that the phase shift of the fast
467: oscillating asymptotics remains to be undefined. Its
468: evaluation within the framework of our approach requires an
469: evaluation on an explicit form of the asymptotics of the
470: higher corrections in the internal expansions at their
471: matching with the fast oscillating asymptotics.
472: \end{demo}
473:
474:
475: \section{Construction of an algebraic asymptotics}
476:
477: In this section the formal asymptotic solution for the
478: equation (\ref{sh}) has been constructed in the form of
479: series on integer powers of the small parameter $\ve$. We show
480: that this expansion is valid in the
481: interval $(t - t_*)\ve^{-4/5} \gg 1$.
482: \par
483: The coefficients of this asymptotics are defined from the
484: recurrent sequence of algebraic equations. These equations
485: can be obtained by substituting of the series (\ref{alas}) in
486: the equation (\ref{sh}) and by collecting of the terms under
487: the same powers of the small parameter. In particular, the
488: algebraic equation (\ref{alg}) for the leader term of the
489: asymptotics can be obtained from the relation under $\ve^0$.
490: \par
491: Further we will use the root $U_2(t)$ as the main term
492: $\stackrel{0}{U}(t)$ of the asymptotics (\ref{alas}). It means
493: that we investigate the formal asymptotics (\ref{alas})
494: corresponding to the stable center which converges with the
495: saddle at $t_*$.
496: \par
497: The relation under $\ve^1$ give us the equations for real and
498: imaginary parts of the first correction term
499: $\stackrel{1}{U}=u_1 +i v_1$
500: \bb
501: [3\stackrel{0}{U}\!^2-t]u_1=0,\quad
502: [\stackrel{0}{U}\!^2-t]v_1=-\pt_t \stackrel{0}{U}. \label{1st}
503: \ee
504: The expression $[3\stackrel{0}{U}\!^2-t]$ vanishes at
505: $t=t_*$. So it is easy to obtain that the even corrections
506: are real, and the odd corrections are imaginary.
507: \par
508: The explicit form of the first correction is:
509: \bb
510: \stackrel{1}{U} = - {i\,{\stackrel{0}{U}}\over
511: {[3\stackrel{0}{U}\!^2-t][\stackrel{0}{U}\!^2-t]}}.
512: \ee
513: This expression has a singularity at $t=t_*$. The order of
514: the singularity increases with respect to the number of the
515: corrections of the asymptotics as $t \to t_*$. In the
516: neighborhood of the point $t_*$ the coefficients of
517: asymptotics (\ref{alas}) have the form:
518: \par
519: in the case of $n=0,1$:
520: \bb
521: \stackrel{n}{U}(t)=(t-t_*)^{-n/2}\sum_{k=0}^\infty\stackrel{n}{U}_k
522: (t-t_*)^{k/2}, \label{expU01} \ee
523: where $\stackrel{0}{U}_0=U_*$,
524: $\stackrel{0}{U}_1=1/\sqrt{3}$;
525: \par
526: in the case of $m=2n,\,\, m=2n+1$, $n\in\Natural$:
527: \bb
528: \stackrel{m}{U}(t)=(t-t_*)^{(1-5m)/2}\sum_{k=0}^\infty\stackrel{m}{U}_k
529: (t-t_*)^{k/2}. \label{expUnm}
530: \ee
531: \par
532: The increasing of the singularity of the corrections in
533: (\ref{alas}) makes this expansion unusable as $t\to t_*$. The
534: singularities of the coefficients define the domain of the
535: suitability for the formal asymptotic solution (\ref{alas})
536: as $(t - t_*)\ve^{-4/5} \gg 1$.
537:
538: \section{Expansions in bifurcation layer}
539:
540: In this section the formal asymptotic solution for equation
541: (\ref{sh}) has been constructed in the domain $|t-t_*| \ll
542: 1$. Usually the asymptotics of this type is called internal
543: ones \cite{I}.
544:
545:
546: \subsection{Initial interval}
547: \subsubsection{The first internal expansion --- Painlev\'e-1 interval}
548:
549:
550: In the neighborhood of the singularity we will use a new
551: scaling variable $\tau=(t-t_*)\ve^{-4/5}$. This new variable
552: $\tau$ is defined by the structure of external asymptotic
553: solution (\ref{alas}) singularities.
554: \par
555: We will construct the solution of the original equation (\ref{sh})
556: in the form:
557: \bb
558: U=U_* + \ve^{2/5}\alpha(\tau,\ve) + i\ve^{3/5}\beta(\tau,\ve),
559: \label{infs}
560: \ee
561: \par
562: The equations for $\a$ and $\b$ are
563: \begin{eqnarray}
564: \a'+(U_*^2-t_*)\b=-\ve^{2/5}2U_*\a\b\,-\,\ve^{4/5}(\a^2-\t)\b\,-\,\ve^{6/5}\b^3,
565: \nonumber\\
566: \b'-3U_*\a^2+U_*\t=\ve^{2/5}(-\a^3-U_*\b^2+\a\t)\,+\,\ve^{4/5}\a\b^2.
567: \label{eq20}
568: \end{eqnarray}
569: \par
570: The asymptotics of the functions $\a$ and $\b$ as
571: $\t\to\infty$ is known. It can be obtained by substituting
572: (\ref{expU01}) and (\ref{expUnm}) in formula (\ref{alas}) and
573: by decomposing of this expression in the terms of the scaling
574: variable $\t$:
575: \begin{eqnarray}
576: \alpha(\tau,\ve) = {1\over\sqrt{3}}\sqrt{\tau}
577: +\sum_{n=1}^{\infty} \tau^{(1-5n)/2}\stackrel{2n}{U}_0 +
578: \ve^{2/5} \sum_{n=0}^{\infty}\stackrel{2n}{U}_1
579: \tau^{(2-5n)/2} +\nonumber\\
580: \ve^{4/5}\sum_{n=0}^{\infty}\stackrel{2n}{U}_2
581: \tau^{(3-5n)/2}+\dots.
582: \label{asst}
583: \end{eqnarray}
584: \begin{eqnarray}
585: \b(\t,\ve)={1\over\sqrt{\t}}\stackrel{1}{U}_0+
586: \sum_{n=1}^{\infty}\stackrel{2n+1}{U}\!\!\!\!_0\,\,\t^{-(1+5n)/2}\,+\,
587: \ve^{2/5}\sum_{n=0}^\infty
588: \stackrel{2n+1}{U}\!\!\!\!_1\,\,\tau^{-(2-5n)/2} +\dots. \label{asbt}
589: \end{eqnarray}
590: \par
591: It is convenient to construct the solution of the equations
592: (\ref{eq20}) in the form of formal series on powers of the
593: small parameter $\ve$:
594: \bb \alpha(\tau,\ve)=
595: \sum_{n=0}^{\infty}\ve^{2n/5}\stackrel{n}{\alpha}(\tau),\quad
596: \beta(\tau,\ve)=
597: \sum_{n=0}^{\infty}\ve^{2n/5}\stackrel{n}{\beta}(\tau). \label{as20}
598: \ee
599: The leader correction terms are defined by the system of ordinary
600: differential equations
601: \begin{eqnarray}
602: \stackrel{0}{\alpha}\!' + (U_*^2 - t_*)\stackrel{0}{\beta}=0,\\
603: \stackrel{0}{\beta}\!'- 3U_*\stackrel{0}{\alpha}\!\!^2+\tau U_* =0.
604: \end{eqnarray}
605: Let us differentiate the first equation and take into account
606: that $U_*(U_*^2-t_*)=1$ then we obtain the Painlev\'e-1
607: equation for $\stackrel{0}{\alpha}$
608: \bb
609: \stackrel{0}{\alpha}{}^{\prime\prime} -
610: 3\stackrel{0}{\alpha}\!\!^2+\tau = 0. \label{p1}
611: \ee
612: Matching the asymptotics (\ref{infs}), (\ref{as20}) and
613: asymptotics (\ref{alas}) as $t\to t_*+0$ we obtain the
614: asymptotic structure of $\stackrel{0}{\alpha}(\tau)$ as $\tau
615: \to +\infty$
616: \bb
617: \stackrel{0}{\alpha}(\tau) = \sum_{n\ge0} \alpha_n
618: \tau^{-{{(5n-1)}\over 2}},\quad{\hbox{where}}\,\,\,\,
619: \a_0={1\over\sqrt{3}},\,\,\a_1={1\over24}. \label{asp1}
620: \ee
621: The solution of the Painleve-1 equation with the asymptotics
622: of (\ref{asp1}) was investigated in \cite{H-S,G-L}.
623: \par
624: The coefficients of the formal expansion (\ref{as20}) for the function
625: $\b$ can be obtained from the first equation of the system (\ref{eq20}).
626: In particular:
627: $$
628: \stackrel{0}{\b}={\stackrel{0}{\a}\!'\over2U_*^2},\quad
629: \stackrel{1}{\b}={1\over2U_*^2}\bigg(\stackrel{1}{\a}\!'+2U_*\stackrel{0}{\a}
630: \stackrel{0}{\b}\bigg),
631: $$
632: $$
633: \stackrel{2}{\b}={1\over2U_*^2}\bigg(\stackrel{2}{\a}\!'+
634: \stackrel{0}{\a}\!^2\stackrel{0}{\b}+2U_*\stackrel{1}{\a}\stackrel{0}{\b}+
635: 2U_*\stackrel{0}{\a}\stackrel{1}{\b}-\t\stackrel{0}{\b} \bigg).
636: $$
637: \par
638: The equation for the first correction term $\stackrel{1}{\alpha}(\tau)$
639: has the form
640: $$
641: \stackrel{1}{\alpha}{}^{\prime\prime} +
642: 6U_*(U_*^2-t_*)\stackrel{0}{\alpha} \stackrel{1}{\alpha}
643: +2U_*(U_*^2-t_*)\stackrel{0}{\beta}\!{}^2 +
644: (U_*^2-t_*)\stackrel{0}{\alpha}\!{}^3
645: -(U_*^2-t_*)\stackrel{0}{\alpha}\tau =
646: $$
647: \bb
648: -2U_*\stackrel{0}{\alpha}{}^\prime\stackrel{0}{\beta} -
649: 2U_*\stackrel{0}{\alpha}\stackrel{0}{\beta}{}^\prime. \label{al1}
650: \ee
651: The asymptotics of a solution for this equation is uniquely
652: defined by matching of the asymptotics of the external
653: expansion (\ref{alas}) and internal expansion (\ref{infs}),
654: (\ref{as20}):
655: \bb
656: \stackrel{1}{\alpha} = \stackrel{0}{U}_2\tau -
657: \stackrel{2}{U}_6\tau^{-3/2} + O(\tau^{-4}), \quad \tau \to +\infty
658: \label{a1-as+}
659: \ee
660: A simple calculation gives us the domain of a suitability of
661: that asymptotic expansion. The representation (\ref{as20}) of
662: the solution is valid as $\tau \ll \ve^{-4/5}$. This result
663: is similar to \cite{Hab1}.
664: \par
665: It is known that the solution of the Painlev\'e-1 equation
666: with the asymptotics (\ref{asp1}) has poles at points $\tau_k$
667: \cite{Kit}. Let us denote the first pole by $\tau_0$ and call
668: it the point of the bifurcation.
669: \par
670: The solution of the Painlev\'e-1 equation is represented by
671: converging series in the neighborhood of the point $\tau_0$
672: \cite{G-L}:
673: $$
674: \stackrel{0}{\alpha}(\tau)=-{2\over{(\tau-\tau_0)^2}} -
675: {\tau_0\over10}(\tau-\tau_0)^2 + a_4(\tau-\tau_0)^4 +
676: \sum_{k=6}^\infty\stackrel{0}{\a}_k(\tau-\tau_0)^k, $$
677: the constants $\tau_0$ and $a_4$ are parameters of the solution.
678: \par
679: The first correction term $\stackrel{1}{\a}$ is defined by
680: the equation (\ref{al1}) and asymptotics (\ref{a1-as+}). In the neighborhood
681: of the bifurcation point $\t_0$ as $\t>\t_0$ the function
682: $\stackrel{1}{\a}$ is represented in the form:
683: \bb
684: \stackrel{1}{\a}=\stackrel{1}{\a}_c(\t)+\stackrel{1}{a}\!^+
685: \a_1(\t)+\stackrel{1}{b}\!^+\a_2(\t). \label{a1-as-bif}
686: \ee
687: Where $\stackrel{1}{\a}_c(\t)$ is the partial solution
688: of the nonhomogeneous equation for the function
689: $\stackrel{1}{\a}(\t)$ with the asymptotics:
690: $$
691: \stackrel{1}{\alpha}_c(\tau) =
692: \stackrel{1}{\a}_{-4}(\tau-\tau_0)^{-4} +
693: \sum_{k=0}^\infty\stackrel{1}{\a}_k(\t-\t_0)^k, \quad
694: \stackrel{1}{\a}_{-4}={1\over {U_*^7}},\quad \stackrel{1}{\a}_4=0.
695: $$
696: It is convenient to use the partial solution with the
697: coefficients as $(\t-\t_0)^{-3}$ and $(\t-\t_0)^4$ equal to zero.
698: \par
699: The constants $\stackrel{1}{a}\!^+$ and $\stackrel{1}{b}\!^+$ in
700: the formula (\ref{a1-as-bif}) can be uniquely defined by
701: asymptotics of $\stackrel{1}{\a}$ as $\t\to\infty$. The functions
702: $\a_1(\t)$ and $\a_2(\t)$ are linear independent solutions of the
703: homogeneous equation for $\stackrel{1}{\a}(\t)$. These functions
704: are uniquely defined by their asymptotics as $\t \to \t_0$:
705: $$
706: \a_1(\t)=(\t-\t_0)^{-3}+a_1(\t-\t_0)+a_2(\t-\t_0)^2+
707: a_3(\t-\t_0)^3+O((\t-\t_0)^5),
708: $$
709: where $a_n=\const$;
710: $$
711: \a_2(\t)=(\t-\t_0)^4+O((\t-\t_0)^8).
712: $$
713: \par
714: The next correction terms are represented by the following
715: formula:
716: $$
717: \stackrel{n}{\a}(\t)=\stackrel{n}{\a}_c(\t)+\stackrel{n}{a}\!^+\a_1(\t)
718: +\stackrel{n}{b}\!^+\a_2(\t).
719: $$
720: The partial solution $\stackrel{n}{\a}_c(\t)$ of the equation has
721: the asymptotics as $\t \to \t_0$:
722: $$
723: \stackrel{n}{\a}_c(\t)=\stackrel{n}{\a}_{-2n}(\t-\t_0)^{-2n}+\dots,
724: $$
725: where dots mean the terms of the next orders on $(\t-\t_0)$.
726: \par
727: It is convenient for us to denote the partial solution which
728: does not contain terms of the order of $(\t-\t_0)^{-3}$ and
729: $(\t-\t_0)^{4}$ in the asymptotics as $\t \to \t_0$ by
730: $\stackrel{n}{\a}_c(\t)$.
731: \par
732: By analysis of written asymptotics we obtain that the formal
733: asymptotic solution (\ref{infs}) is valid as
734: $(\tau - \tau_0)\ve^{-1/5} \gg 1$.
735:
736: \subsubsection{The second internal expansion, the neighborhood
737: of the bifurcation point}
738:
739: In this section we construct the formal asymptotic solution
740: for the equation (\ref{sh}), which is valid in the
741: neighborhood of the bifurcation point $|\tau - \tau_0| \ll 1$.
742: \par
743: The domain where the first asymptotic expansion is valid
744: determines the second internal scale. We will use a new
745: scaling variable $\th=(\tau - \tau_0)\ve^{-1/5}$ in the small
746: neighborhood of the point $\tau_0$.
747: \par
748: We change the unknown function in the equation (\ref{sh}) and pass to
749: the function $w=U-U_*$.
750: \par
751: Substituting the expressions for $w$ and $\th$ in the original
752: problem we obtain the equation for $w(\th,\ve)$
753: \begin{eqnarray}
754: i w' +2U_*^2w+2U_*|w|^2+|w|^2w+U_*^2 w^* +U_* w^2- \nonumber\\
755: -\ve^{4/5}\tau_0 \big(U_*+w\big)-\ve\th\big( U_*+w\big)
756: =0.\label{w-main}
757: \end{eqnarray}
758: \par
759: The asymptotics of the function $w(\th,\ve)$ as $\th\to
760: +\infty$ can be obtained by expansing in series of the
761: function $U=U_*+\ve^{2/5}\a(\t,\ve)+i\ve^{3/5}\b(\t,\ve)$ on
762: variable $\th$:
763: \begin{eqnarray}
764: w(\th,\ve)=\bigg({1\over U_*^3\th^2}+{i\over2U_*^5\th^3} +\dots\bigg)
765: +\ve^{1/5}\stackrel{1}{a}\!^+\bigg(\th^{-3}+\dots\bigg)+
766: \nonumber\\
767: \ve^{3/5}\stackrel{2}{a}\!^+\bigg(\th^{-3}+\dots\bigg) +
768: \ve^{4/5}\bigg({2 U_8^3\t_0\over10}\th^{2}+\dots\bigg)+\dots.
769: \label{w-as-right}
770: \end{eqnarray}
771: In this formula dots mean the terms of the lesser order (in
772: brackets with respect to $\th$ and out of them with respect
773: to $\ve$).
774: \par
775: In this section the formal asymptotic solution for the
776: equation (\ref{w-main}) with the given asymptotics
777: (\ref{w-as-right}) as $\ve\to0$ has been constructed. We
778: indicate the domain where this formal solution for the
779: equation (\ref{w-main}) is valid.
780: \par
781: The first step of the formal solution construction for the
782: equation (\ref{w-main}) is a choice of an asymptotic
783: sequence. This choice for the external and the first internal
784: expansions was defined by the equation. In this case the
785: asymptotics as $\th\to+\infty$ gives us the asymptotic
786: sequence on $\ve^{1/5}$. But the equation shows, that the
787: first order correction should satisfy to the homogeneous
788: equation. In this case we can satisfy to the asymptotic
789: condition as $\th\to+\infty$ by constructing of the formal
790: solution in the form:
791: \bb
792: w(\th,\ve)=\stackrel{0}{w}(\th_0)\,+\,
793: \ve^{4/5}\sum_{n=1}^\infty\ve^{(n-1)/5}\stackrel{n}{w}(\th_0),
794: \label{exp-w} \quad \ve\to0
795: \ee
796: where
797: $$
798: \th_0=\th+\ve^{1/5}\stackrel{1}{\th_0}+
799: \ve^{3/5}\stackrel{2}{\th_0}+\dots \quad \ve\to0.
800: $$
801: The correction terms $\stackrel{n}{\th}$ are defined by the
802: asymptotics of the formal solution as $\th\to\infty$:
803: $$
804: \stackrel{n}{\th_0}={\stackrel{n}{a}\!\!^+\over4}. $$
805: \par
806: It is convenient to pass from the equation (\ref{w-main}) to
807: the equation with respect to the variable $\th_0$:
808: \begin{eqnarray}
809: iw' +(2U_*^2-t_*)w+2U_*|w|^2+|w|^2w+U_*^2 w^* +U_* w^2=\nonumber\\
810: \bigg(\ve^{4/5}\tau_0+\ve\th_0 -\ve^{6/5}\stackrel{1}{\th}_0
811: -\ve^{8/5}\stackrel{2}{\th}_0-\dots\bigg)\big( U_*+w\big),
812: \quad\ve\to0.
813: \label{w-main0}
814: \end{eqnarray}
815: \par
816: To construct the formal solution $w(\th_0,\ve)$ we substitute
817: the formal series (\ref{exp-w}) in the equation
818: (\ref{w-main0}) and gather the terms of the same order on
819: $\ve$. Then we obtain the recurrent sequence of the equations
820: for the leader term and the corrections of the asymptotics.
821: The equation for the leader term is
822: \bb
823: i\stackrel{0}{w}{}^\prime +
824: U_*\left(2|\stackrel{0}{w}|^2+\stackrel{0}{w}\!{}^2\right) + U_*^2
825: \left( \stackrel{0}{w}\!^* - \stackrel{0}{w} \right) +
826: |\stackrel{0}{w}|^2 \stackrel{0}{w}= 0. \label{w0}
827: \ee
828: \par
829: The asymptotics of $\stackrel{0}{w}$ as $\th_0\to\infty$ can be obtained
830: from the asympotics of the function $w(\th_0,\ve)$
831: \bb
832: \stackrel{0}{w}(\th_0)={1\over
833: U_*^3\th_0^2}+{i\over2U_*^5\th_0^3}+\dots. \label{w0-as-right} \ee
834: \par
835: To construct the leader term $\stackrel{0}{w}(\th_0)$ we take
836: the conservation law for the equation (\ref{w0}):
837: $$
838: U_*\left[|w|^2w^* + |w|^2w \right] + U_*^2\left[{1\over2} (w^*)^2 +
839: {1\over2} w^2 -|w|^2 \right] + {1\over2}|w|^4 = H.
840: $$
841: The constant $H$ for our solution is defined by asymptotics
842: (\ref{w0-as-right}): $H=0$. Let us solve this formula for the
843: function $w^*$ and substitute the representation over $w$
844: in the equation (\ref{w0}). Then we obtain
845: \bb
846: iw^{\prime} - 2U_*w\sqrt{-U_*w}=0. \label{mainwa}
847: \ee
848: This equation can be obviously solved
849: \bb
850: \stackrel{0}{w}(\th_0) ={{-2}\over{(\th_0-iU_*)^2}}, \label{wa}
851: \ee
852: the integration constant is defined by the asymptotics
853: (\ref{w0-as-right}). The leader term of the asymptotics
854: (\ref{exp-w}) is the separatrix solution of the equation
855: (\ref{w0}).
856: \par
857: The next corrections $\stackrel{m}{w},\,\,m=1,2 \dots$
858: satisfy to the equations
859: \bb i\stackrel{m}{w}{}^\prime
860: +\left[2U_*\stackrel{0}{w}\!^* + 2U_*\stackrel{0}{w} - U_*^2 + 2
861: \stackrel{0}{w} \stackrel{0}{w}\!^* \right]\stackrel{m}{w} +
862: \left(U_* + \stackrel{0}{w}\right)^2 \stackrel{m}{w}\!^* =
863: \stackrel{m}{F}. \label{mwa} \ee
864: In particular,
865: $$
866: \stackrel{1}{F}=\tau_0 (U_* + \stackrel{0}{w}),\quad
867: \stackrel{2}{F}=\theta_0(U_*+\stackrel{0}{w})\quad
868: \stackrel{3}{F}=-\stackrel{1}{\theta_0}(U_*+\stackrel{0}{w}),
869: \quad \stackrel{4}{F}=0, $$ $$
870: \stackrel{5}{F}=-\stackrel{2}{\theta}(U_*+\stackrel{0}{w})-
871: 2U_*|\stackrel{1}{w}|^2-U_*\stackrel{1}{w}\!\!^2. $$
872: The asymptotics $\stackrel{m}{w}(\th_0)$ as $\th\to\infty$
873: can be easy obtained from (\ref{w-as-right}).
874: \par
875: To construct the correction terms we need the solutions of the homogeneous
876: equation (\ref{mwa}). One of them can be obtained by differentiation
877: with respect to variable $\th_0$ of the function $\stackrel{0}{w}$
878: \bb
879: w_1(\th_0)={1\over 4}{d \over {d\th_0}}\stackrel{0}{w} =
880: {1\over(\th_0-iU_*)^3}.\label{w1}
881: \ee
882: \par
883: The formula for the second independent solution is complicated
884: and we do not write it. The second solution grows as $|\th|\to\infty$.
885: The asymptotics as $\th\to\infty$ of this solution has
886: the form
887: \bb w_2(\th_0)=\th_0^4+
888: i{1 \over {2U_*^2}}\th_0^3+\dots.\label{w2}
889: \ee
890: These independent solutions are such that their
891: Wronskian is
892: $$
893: {\cal W}=w_1w_2^*-w_2w_1^*=-{7i\over U_*^2}.
894: $$
895: \par
896: By using the expressions for $w_1$ and $w_2$ we can write
897: the next corrections for the asymptotic solution with
898: the given asymptotics as $\th_0\to+\infty$ in an
899: explicit form. The asymptotics as
900: $\th_0\to-\infty$ of these corrections can be evaluated.
901: \par
902: It is convenient for us to represent the higher corrections of the
903: asymptotics (\ref{exp-w}) in the form
904: \bb
905: \stackrel{n}{w}(\th_0)=\stackrel{n}{w}_c(\th_0)+\stackrel{n}{X}_0\!\!^-w_1(\th_0)+
906: \stackrel{n}{Y}_0\!\!^-w_2(\th_0), \label{1waa} \ee where,
907: $\stackrel{n}{w}_c(\th_0)$ is the partial solution for the n-th
908: correction term and this solution does not contain terms of the order
909: of $\th_0^4$ and $\th_0^{-3}$ in the asymptotics as $\th_0=\infty$.
910: The constants $\stackrel{n}{X}_0\!\!^-$ and $\stackrel{n}{Y}_0\!\!^-$
911: can be evaluated by
912: $$
913: \stackrel{n}{X}_0\!\!^-=-\res_{r=\infty}\bigg[\,{1\over
914: r}\,\int_{-r}^{r}\big(\stackrel{n}{F}w_2^*+
915: \stackrel{n}{F}\!\!^*w_2\big){d\th\over{\cal W}}\bigg], $$ $$
916: \stackrel{n}{Y}_0\!\!^-=-\res_{r=\infty}\bigg[\,{1\over
917: r}\,\int_{-r}^{r}\big(\stackrel{n}{F}w_1^*+
918: \stackrel{n}{F}\!\!^*w_1\big){d\th\over{\cal W}}\bigg]. $$
919: \par
920: The segment of the Laurent series for the function
921: $\stackrel{1}{w}_c(\th_0)$ has the form
922: $$
923: \stackrel{1}{w}_c(\th_0) = {{\tau_0 U_*^3}\over 5}\th_0^2 +
924: i{{\tau_0 U_*}\over 5}\th_0 - {\tau_0\over{30U_*}} +
925: i{{3\tau_0}\over{5U_*^3}}\th_0^{-1} + O(\th_0^{-2}), \quad \th
926: \to\infty. $$
927: The coefficients as $w_1$ and $w_2$ are
928: $$
929: \stackrel{1}{X}_0\!\!^-=0,\quad \stackrel{1}{Y}_0\!\!^-=0. $$
930: \par
931: The function $\stackrel{2}{w}(\th_0)$ as $\th_0\to-\infty$
932: also can be represented in the form of sum of the partial solution
933: of the nonhomogeneous equation and solutions of the homogeneous
934: equation
935: \bb
936: \stackrel{2}{w}(\th_0)=\stackrel{2}{w}_c(\th_0)+\stackrel{2}{X}_0\!\!^-w_1(\th_0)+
937: \stackrel{2}{Y}_0\!\!^-w_2(\th_0). \label{as2ct}
938: \ee
939: The partial solution of the nonhomogeneous equation can
940: be represented in the form as $\th_0\to-\infty$
941: $$
942: \stackrel{2}{w}_c(\th_0)=-{\th_0^3\over6}+i{U_*\th_0^2\over2}+{5\th_0\over12
943: U_*}-i{3\over4}+
944: {15U_*\over16\th_0}+i\big({16U_*^2\over15}+{2\over3U_*}\big)\th_0^{-2}+O(\th_0^{-4}).
945: $$
946: The Laurent series of this solution does not contain the terms of the order
947: of $\th_0^4$ and $\th_0^{-3}$. The coefficients as
948: $w_1$ and $w_2$ are
949: $$
950: \stackrel{2}{X}_0\!\!^-=0,\quad \stackrel{2}{Y}_0\!\!^-={\pi\over 2
951: }$$
952: \par
953: The next correction terms of the asymptotics (\ref{exp-w}) also can be
954: obtained by this way. The condition
955: $\ve\stackrel{2}{w}/(\ve^{4/5}\stackrel{1}{w})\ll1$ gives us the domain
956: where the expansion (\ref{exp-w}) of the asymptotic solution is valid
957: $$ -\ve^{-1/5}\ll\th_0\ll\ve^{-1/10}. $$
958: \begin{demo}
959: It is easy to see that the asymptotics of (\ref{exp-w}) as
960: $\th_0\to\infty$ and as $\th_0\to-\infty$ are different. The
961: difference is the term of the order of $\ve\th^4$. It means
962: that additional terms appear via the passage of the
963: neighborhood of the bifurcation point. The leader term of
964: these additional terms is $\stackrel{2}{Y}\!\!^-_{0}=\ve
965: \pi\th^4/2$. This effect leads to the appearance of the
966: solution $\stackrel{2}{Y}\!\!^-_0 w_2$ of the homogeneous
967: equation in the asymptotics of the second corrections as
968: $\th_0\to-\infty$.
969: \end{demo}
970:
971:
972: \subsubsection{Intermediate expansion, initial interval}
973:
974: In this section we investigate the behaviour of the solution
975: after a passage of the narrow neighborhood of the bifurcation
976: point $\tau_0$.
977: \par
978: In this domain it is convenient
979: to introduce a new scaling variable $T_1=\th_0\ve^{1/6}$ by
980: analysis of the second internal expansion.
981: \par
982: The solution of the equation (\ref{sh}) we seek in the form:
983: \bb
984: U=U_* + \ve^{2/6} A(T_1,\ve) + i \ve^{3/6}B(T_1,\ve),
985: \label{3a}
986: \ee
987: where $A(T_1,\ve)$ and $B(T_1,\ve)$ are real functions.
988: \par
989: The equation (\ref{sh}) can be written in terms of these new
990: variables in the form:
991: \begin{eqnarray}
992: A'-2U_*^2 B=-\ve^{1/3}2U_*AB-\ve^{2/3}A^2B-\ve B^3+\nonumber
993: \\
994: \big(\ve^{4/5}\t_0+
995: \ve^{5/6}T_1-\ve^{4/5}\sum_{n=1}^{\infty}\ve^{(2n-1)/5}\stackrel{n}{\th}_0
996: \big)B, \nonumber
997: \\
998: B'+3U_* A^2=-\ve^{1/3}(A^3-B^2U_*)-\ve^{2/3}AB^2+
999: \big(\ve^{1/6}T_1+\ve^{2/15}\t_0-
1000: \nonumber\\
1001: \ve^{2/15}\sum_{n=1}^\infty\ve^{(2n-1)/5}\stackrel{n}{\th}_0\big)
1002: \big(U_*+\ve^{1/3}A\big),\ve\to0.
1003: \label{3ineq}
1004: \end{eqnarray}
1005: \par
1006: The asymptotics of $U$ as $\th\to-\infty$ rewriting in terms of
1007: the variable $T_1$ gives the asymptotics of $A$ and $B$ as
1008: $T_1\to-0$:
1009: \begin{eqnarray}
1010: A=\bigg({-2\over T_1^2}+\stackrel{2}{Y}\!\!^-_0
1011: T_1^4+\dots\bigg)+
1012: \ve^{2/15}\bigg({\t_0U_*^3\over5}T_1^2+\dots\bigg)+\nonumber\\
1013: \ve^{1/6}\bigg({-1\over6}T_1^3+\dots\bigg)+\dots,\nonumber\\
1014: B=\bigg({-2\over U_*^2T_1^3}+\stackrel{2}{Y}\!^-_0T_1^3+\dots\bigg)+
1015: \ve^{2/15}\bigg({\t_0U_*\over5} T_1+\dots
1016: \bigg)+\dots.\label{AB-as-left}
1017: \end{eqnarray}
1018: The dots in the brackets mean the terms of the less order with
1019: respect to $T_1$ and out of the brackets -- with respect to
1020: $\ve$.
1021: \par
1022: The system (\ref{3ineq}) and the asymptotics
1023: (\ref{AB-as-left}) contain terms of the different orders with
1024: respect to $\ve$. It is convenient to represent the formal
1025: asymptotic solution with respect to $\ve$ in the form:
1026: \begin{eqnarray}
1027: A(T_1,\ve) = \sum_{n=0}^\infty\ve^{n/30}\stackrel{n}{A}(T_1), \nonumber\\
1028: B(T_1,\ve) =
1029: \sum_{n=0}^\infty\ve^{n/30}\stackrel{n}{B}(T_1),\quad \ve\to0.
1030: \label{AB-anzatz}
1031: \end{eqnarray}
1032: \par
1033: By substituting the representation for $A$ and $B$ in the
1034: equation we obtain the recurrent sequence of the equations for
1035: $\stackrel{n}{A}(T_1),\ \stackrel{n}{B}(T_1)$. It is
1036: convenient for us to investigate the second order
1037: differential equation for the functions
1038: $\stackrel{n}{A}(T_1)$ instead of the system of differential
1039: equations of the first order for
1040: $\stackrel{n}{A}(T_1),\,\stackrel{n}{B}(T_1)$. The equation
1041: for the leader term is
1042:
1043: \bb
1044: \stackrel{0}{A}\!\!^{\prime\prime} +
1045: 3\stackrel{0}{A}\!^2 = 0. \label{glp}
1046: \ee
1047: The equations for the next corrections of the asymptotics
1048: (\ref{AB-anzatz}) are
1049: $$
1050: \stackrel{m}{A}\!''+6\stackrel{0}{A}\stackrel{m}{A}=\stackrel{m}{F},
1051: $$
1052: in particular,
1053: \begin{eqnarray*}
1054: \stackrel{m}{F}\equiv0,\,\,m=1,2,3,6,7;\quad
1055: \quad\stackrel{4}{F}=-2U_*^2\t_0;\quad\stackrel{5}{F}=-2U_*^2T_1;\\
1056: \stackrel{8}{F}=-3\stackrel{4}{A}\!\!^2(T_1);\quad\quad\quad
1057: \stackrel{9}{F}=-6\stackrel{4}{A}(T_1)\stackrel{5}{A}(T_1);
1058: \\ \stackrel{10}{F}=-\stackrel{1}{\th}_0
1059: U_*-2U_*\bigg({\big(\stackrel{0}{A}\!\!'\big)^2\over2U_*}\,+
1060: \,{3\stackrel{0}{A}\over2U_*^2}\bigg)-2U_*^2\bigg(\stackrel{0}{A}\!\!\!^3-
1061: {\big(\stackrel{0}{A}\!\!'\big)^2\over4U_*^3}\bigg).
1062: \end{eqnarray*}
1063: \par
1064: The asymptotics as $T_1\to-0$ of the coefficients of the formal
1065: series for $A(T_1,\ve)$ can be obtained from the formulas
1066: (\ref{AB-as-left})
1067: \begin{eqnarray*}
1068: \stackrel{0}{A}={-2\over T_1^2}+ \stackrel{2}{Y}\!\!^-_0
1069: T_1^4+\dots;\quad
1070: \stackrel{m}{A}\equiv0,\,\,m=1,2,3,6,7;\\
1071: \stackrel{4}{A}={\t_0U_*^3\over5}T_1^2+\dots;\quad
1072: \stackrel{5}{A}={-1\over6}T_1^3+\dots;\\
1073: \stackrel{8}{A}(T_1)=O(T_1^6);\quad \stackrel{9}{A}(T_1)=O(T_1^7);
1074: \quad \stackrel{10}{A}(T_1)=O(T_1^{-4}).
1075: \end{eqnarray*}
1076: The dots in the formulas mean the terms of the less order with
1077: respect to $T_1$. In particular, these formulas give us the
1078: domain where the segment of the length $n=10$ of the formal
1079: series (\ref{AB-anzatz}) as $T_1\to-0$ is valid
1080: $$
1081: {\ve^{10/30}\stackrel{10}{A}(T_1)\over\stackrel{0}{A}(T_1)}\ll1,\,\,\,
1082: {\hbox{or}}\,\,\, T_1\gg\ve^{1/6}.
1083: $$
1084: \par
1085: Let us construct the formal asymptotic solution in the form
1086: (\ref{AB-anzatz}). The seeking solution
1087: $\stackrel{0}{A}(T_1)$ for the leader term equation
1088: (\ref{glp}) has the asymptotics as $T_1\to -0$ in the form
1089: $$
1090: \stackrel{0}{A}=-{2\over T_1^2}+a_4(1) T_1^4+O(T_1^{10}),
1091: $$
1092: where $a_4(1)=\stackrel{2}{Y}\!\!^-_0$. Using this
1093: asymptotics it is easy to see that the function
1094: $\stackrel{0}{A}$ is expressed by the Weierstrass function
1095: $$
1096: \stackrel{0}{A}(T_1)=-2\wp(T_1;0,g_3(1)),\quad{\hbox{where}}\quad
1097: g_3(1)=a_4(1)/56.
1098: $$
1099: \par
1100: In a general case the solution $\stackrel{n}{A}(T_1)$ can be
1101: represented in the neighborhood of the point $T_1=0$ in the
1102: form
1103: $$
1104: \stackrel{n}{A}(T_1)=\stackrel{n}{A}_0\!\!\!^c(T_1) +
1105: \stackrel{n}{x}_0\!\!\!^- A_1(T_1)+ \stackrel{n}{y}_0\!\!\!^-
1106: A_2(T_1).
1107: $$
1108: The function $\stackrel{n}{A}_0\!\!\!^c(T_1)$ is the partial
1109: solution of the nonhomogeneous equation for the $n$-th correction.
1110: This solution does not contain terms of the order of $T_1^{-3}$ and $T_1^4$.
1111: The functions $A_1(T_1)$ and $A_2(T_1)$ are linear independent
1112: solutions of the linear homogeneous equation in variations.
1113: These solutions are uniquely defined by their asymptotics as
1114: $T_1\to-0$
1115: $$
1116: A_1(T_1)=T_1^{-3}+\dots,\quad
1117: A_2(T_1)=T_1^4+\dots,
1118: $$
1119: the expansion of the function $A_1(T_1)$ does not contain
1120: terms of the order of $T_1^4$. Both of these solutions can be represented
1121: via the Weierstrass function. In the case $g_2=0$
1122: we obtain (\cite{Abramowitz})
1123: $$
1124: A_1(T_1)={1\over4}\wp'(T_1,0,g_3),
1125: $$ $$ A_2(T_1)=-14\pt_{g_3}\wp(T_1,0,g_3)\equiv
1126: {7\over3g_3}T_1\wp'(T_1,0,g_3)+{14\over3g_3}\wp(T_1,0,g_3). $$
1127: The Wronskian of these solutions is $ W=-{7\over4}$.
1128: \par
1129: The coefficients $\stackrel{n}{x}_0\!\!\!^-$ É
1130: $\stackrel{n}{y}_0\!\!\!^-$ are defined by matching of the asymptotics
1131: (\ref{AB-anzatz}) as $T_1\to-0$ and the asymptotics (\ref{exp-w}) as
1132: $\th_0\to\infty$
1133: $$
1134: \stackrel{n}{x}_0\!\!\!^-=\stackrel{m}{X}_0\!\!\!^-,\quad n=6m+23;
1135: $$ $$ \stackrel{n}{y}_0\!\!\!^-=\stackrel{m}{Y}_0\!\!\!^-,\quad
1136: n=6m-12. $$
1137: \par
1138: The function $\stackrel{0}{A}$ has poles. One of them is
1139: $T_1=0$. The second pole of this function is defined by an
1140: elliptic integral
1141: $$
1142: \O_1=2\int_{A_{min}}^\infty{dy\over\sqrt{4y^3-g_3(1)}}, \quad
1143: A_{min}=\bigg({g_3(1)\over4}\bigg)^{1/3}. $$
1144: \par
1145: The formal asymptotics (\ref{AB-anzatz}) loses the asymptotic
1146: behaviour in the neighborhoods of poles. The asymptotics as
1147: $T_1\to-0$ is known. These explicit formulas give us the
1148: domain where the asymptotic series (\ref{AB-anzatz}) is valid
1149: as $-T_1\gg\ve^{1/6}$.
1150: \par
1151: To construct the asymptotics of the coefficients of the formal
1152: solution as $T_1\to-\O_1+0$ we solve the equations for the correction
1153: terms and write their expansions as $T_1\to-\O_1+0$.
1154: \par
1155: It is convenient to represent the correction $\stackrel{n}{A}$ as
1156: $T_1\to-\O_1+0$ in the form
1157: \bb
1158: \stackrel{n}{A}(T_1)=\stackrel{n}{A}_c(T_1)+
1159: \stackrel{n}{x}_1\!\!\!^+ A_1(T_1+\O_1)+ \stackrel{n}{y}_1\!\!\!^+
1160: A_2(T_1+\O_1). \label{n-correction}
1161: \ee
1162: The function $\stackrel{n}{A}_c(T_1)$ is the solution of the nonhomogeneous
1163: equation for the $n$-th correction. The asymptotics as $T_1 \to -\O_1+0$ of
1164: this function does not contain terms of the order of $(T_1+\O_1)^{-3}$ and
1165: $(T_1+\O_1)^4$.
1166: \par
1167: The constants in the formula for the $n$-th correction are evaluated
1168: by formulas
1169: $$
1170: \stackrel{n}{y}_1\!\!\!^+=\stackrel{n}{y}_0\!\!\!^-\,+\,\res_{r=0}\bigg[{1\over
1171: r} \int_{-r}^{-\O_1+r} {dz\,\stackrel{n}{F}(z)A_1(z)\over W}
1172: \bigg],
1173: $$
1174: $$
1175: \stackrel{n}{x}_1\!\!\!^+= \stackrel{n}{x}_0\!\!\!^-\,+\,
1176: C_1\stackrel{n}{y}_1\!\!^+\, +\,\res_{r=0}\bigg[{1\over r}
1177: \int_{-r}^{-\O_1+r} {dz\,\stackrel{n}{F}(z)A_2(z)\over W}
1178: \bigg],
1179: $$
1180: where the constant $C_1$ is defined by
1181: $$
1182: A_2(T_1+\O_1)=C_1A_1(T_1)+A_2(T_1).
1183: $$
1184: \par
1185: To obtain the domain of the suitability of the formal
1186: asymptotics as $T_1\to-\O_1+0$ it is necessary to know the
1187: type of the singularity of the correction terms. The first
1188: nonzero correction term is $\stackrel{4}{A}(T_1)$. The
1189: expression (\ref{n-correction}) for $n=4$ contains the
1190: partial solution $\stackrel{4}{A}_1\!\!\!^c(T_1)$. This
1191: function is smooth and has the identical asymptotics as
1192: $T_1\to-\O_1+0$ and as $T_1\to-0$. But the asymptotics of the
1193: $\stackrel{4}{A}$ as $T_1\to-\O_1+0$ contains terms of the
1194: order of $(T_1+\O_1)^4$ with the coefficients
1195: $\stackrel{4}{x}_1\!\!\!^+\not=0$. Namely,
1196: $$
1197: \stackrel{4}{y}_1\!\!\!^+=0,
1198: $$
1199: $$
1200: \stackrel{4}{x}_1\!\!\!^+={28\t_0\over3g_3}\z\big(\O_1/2;0,g_3\big),
1201: $$
1202: where $\z(T_1,g_2,g_3)$ is the Weierstrass $\z$-function.
1203: \par
1204: The values of the coefficients $\stackrel{n}{y}_1\!\!\!^+$ and
1205: $\stackrel{n}{x}_1\!\!\!^+$ for $n>0$ also are evaluated. To
1206: obtain the domain where the asymptotics
1207: (\ref{AB-anzatz}) is valid it is necessary to note
1208: $$
1209: {\ve^{2/15}\stackrel{4}{A}(T_1)\over\stackrel{0}{A}(T_1)}=
1210: O\left({\ve^{2/15}\over (T_1+\O_1)}\right)\ll 1.
1211: $$
1212: It is easy to see that the condition
1213: $\ve^{-2/15}(\O_1+T_1)\gg1$ is the condition of the
1214: suitability of the asymptotics (\ref{AB-anzatz}) as $T_1\to
1215: -\O_1 + 0$.
1216: \par
1217: Thus, we have constructed the formal asymptotic solution of
1218: the original equation in the form (\ref{AB-anzatz}) in the
1219: domain where the following conditions $-\ve^{-1/6}T_1\gg1$
1220: and $\ve^{-2/15}(\O_1+T_1)\gg1$ are valid.
1221:
1222:
1223:
1224:
1225: \subsection{The bifurcation layer in the case of bounded $k$}
1226: \subsubsection{The neighborhoods of the poles $T_k$ of the
1227: Weierstrass function}
1228:
1229: In this section we construct the solution of the original
1230: problem in the neighborhoods of the poles of the solutions of
1231: (\ref{glp}) and show the change of the solution via the
1232: passage from one pole to another one.
1233: \par
1234: In the neighborhood of the poles we will use a new scaling
1235: variable
1236: \bb
1237: \th_k =
1238: (T_k+\O_k-{1\over4}\sum_{n=1}^\infty\ve^{n/30}\stackrel{n}{x}_k\!\!\!^+)
1239: \ve^{-1/6},\quad \ve\to0. \label{thk}
1240: \ee
1241: The original equation (\ref{sh}) looks like
1242: \begin{eqnarray}
1243: iU' + |U|^2U-\left(t_*+\ve^{4/5}\bigg(\tau_0\,+
1244: \,\sum_{n=1}^\infty\ve^{(2n-1)/5}\stackrel{n}{\th}_0\bigg)
1245: \,+ \, \right. \nonumber
1246: \\
1247: \left. \ve^{5/6}\bigg(\sum_{j=1}^k\O_j \,+\,
1248: {1\over4}\sum_{n=1}^{\infty}\ve^{n/30}
1249: \sum_{j=1}^k\stackrel{n}{x}_j\!\!\!^+ \,+\, \ve^{1/6}
1250: \th_k\bigg) \right)U=1,\quad \ve\to0, \label{Uk-eq}
1251: \end{eqnarray}
1252: where $\tau_0$ is the bifurcation point and $\Omega_k$ are poles
1253: of the Weierstrass function which describe the leader
1254: term of the asymptotics in the intermediate domains.
1255: \par
1256: The solution is seeking in the form of formal series
1257: \bb
1258: U=U_* + \stackrel{0}{W}(\th_k) \,+\,
1259: \ve^{4/5}\sum_{n=1}^\infty\ve^{(n-1)/30}\stackrel{n}{W}(\th_k),
1260: \quad \ve\to0.
1261: \label{weia}
1262: \ee
1263: \par
1264: The coefficients of the formal expansion (\ref{weia}) are
1265: satisfied to the recurrent system of equations. In
1266: particular, the function $\stackrel{0}{W}(\th_k)$ is
1267: satisfied to the equation (\ref{w0}). The equations for the
1268: higher corrections of the asymptotics are
1269: \bb
1270: i\stackrel{n}{W}{}^\prime +\left[2U_*\stackrel{0}{W}\!^* +
1271: 2U_*\stackrel{0}{W} - U_*^2 + 2 \stackrel{0}{W}
1272: \stackrel{0}{W}\!^* \right]\stackrel{n}{W} + \left(U_* +
1273: \stackrel{0}{W}\right)^2 \stackrel{n}{W}\!^* = \stackrel{n}{F}.
1274: \label{enwaa}
1275: \ee
1276: \par
1277: Let us become clear how the nonlinearity $\stackrel{n}{F}$
1278: depends on $\stackrel{m}{W}$ and on
1279: $\stackrel{j}{x}_k\!\!\!^+$. The maximum order of nonlinear
1280: terms is cubic. The function $\stackrel{n}{F}$ contains terms
1281: of the type
1282: $$ \stackrel{j}{W}\stackrel{l}{W}\stackrel{m}{W}, \quad
1283: n=j+l+m+46,\,\,j,l,m\ge1; $$
1284: and
1285: $$
1286: (U_*+\stackrel{0}{W})\stackrel{j}{W}\stackrel{l}{W}, \quad
1287: n=j+l+23, \,\,j,l\ge1; $$
1288: linear terms with constant coefficients with dependness on
1289: $\stackrel{j}{x}_l\!\!\!^+$
1290: $$
1291: \stackrel{m}{W}\sum_{l=1}^k \stackrel{j}{x}_l\!\!\!^+,\quad
1292: n=m+j+25,l,\,m,j\ge1; $$
1293: linear terms with the coefficients depending on $\th_k$
1294: $$ \th_k\stackrel{m}{W},\quad n=m+55; $$
1295: and with coefficients only depending on
1296: $\stackrel{m}{x}_j\!\!\!^+$
1297: $$
1298: \sum_{j=1}^k\stackrel{m}{x}_j\!\!\!^+(U_*+\stackrel{0}{W}),\quad
1299: n=m+2; $$ and linear terms with depending on
1300: $\stackrel{m}{\th}_0$:
1301: $$
1302: \stackrel{m}{\th}_0\stackrel{l}{W},\quad n=12m+l+18;
1303: $$
1304: and the terms depending on the leader term of the asymptotics
1305: $$
1306: \stackrel{m}{\th}_0(U_*+\stackrel{0}{W}),\quad n=12m-5.
1307: $$
1308: \par
1309: For example, we write a few of the right hand-side terms for
1310: the lower corrections:
1311: $$
1312: \stackrel{1}{F}=\tau_0 (U_* + \stackrel{0}{W});\quad
1313: \stackrel{2}{F}=(U_* +
1314: \stackrel{0}{W})\sum_{j=1}^k\O_j;\quad\,
1315: \stackrel{n}{F}\equiv0,\,\,n=3,4,5,6; \quad
1316: $$
1317: $$
1318: \stackrel{7}{F}=(U_* +
1319: \stackrel{0}{W})(\th_k+{1\over4}\sum_{j=1}^k\stackrel{5}{x}_j\!\!\!^+).
1320: $$
1321: It is easy to see that the right-hand side of the equations
1322: for a few of the lower corrections are proportional to leader
1323: term of the asymptotics. The right-hand side for the 7-th
1324: correction term depends on the leader term by more complicated
1325: way. There is the variable coefficient $\th_k$ in the
1326: equation.
1327: \par
1328: The solutions of the equations for the corrections terms of
1329: the asymptotics can be represented in the form
1330: \bb
1331: \stackrel{n}{W}(\th_k)=\stackrel{n}{W}_c(\th_k) \,+\,
1332: \stackrel{n}{X}_k\!\!\!^+\,w_1(\th_k) \,+\,
1333: \stackrel{n}{Y}_k\!\!\!^+\,w_2(\th_k).\label{Wn} \ee Where
1334: the function $\stackrel{n}{W}_c(\th_k)$ is the solution of
1335: the nonhomogeneous equation for the $n$-th correction term.
1336: This solution does not contain the terms of the order of
1337: $\th_k^4$ and $\th_k^{-3}$ in its asymptotics as
1338: $\th_k\to\infty$. The linear independent solutions of the
1339: homogeneous equation $w_1(\theta_k)$ and $w_2(\theta_k)$ are
1340: defined in the section 5.1.2. The constants
1341: $\stackrel{n}{X}_k\!\!^+$ and $\stackrel{n}{Y}_k\!\!^+$ will
1342: be determined below.
1343: \par
1344: The asymptotics of the function $\stackrel{0}{W}(\th_k)$ as
1345: $\th_k\to\infty$ is given. It can be obtained by rewriting of the
1346: formal solution $U_*+\ve^{2/6}A(T_k,\ve)+i\ve^{3/6}B(T_k,\ve)$ as
1347: $T_k\to-\O_k+0$ in the terms of the scaling variable $\th_k$.
1348: This asymptotics
1349: uniquely determines the leader term of the formal solution
1350: \bb
1351: \stackrel{0}{W}(\th_k)={-2\over(\th_k-iU_*)^2}. \label{W0}
1352: \ee
1353: Thus, the dependence on the number $k$ in the formula for
1354: the leader term of the formal series is only contained in the
1355: argument of the variable $\th_k$.
1356: \par
1357: To evaluate the constants $\stackrel{n}{X}_k\!\!\!^+$ and
1358: $\stackrel{n}{Y}_k\!\!\!^+$ in the formula for the $n$-th
1359: corrections we match its asymptotics as $\th_k\to\infty$ with
1360: the asymptotics obtained by rewriting the formal series
1361: $U_*+\ve^{2/6}A(T_k,\ve)+i\ve^{3/6}B(T_k,\ve)$ as
1362: $T_k\to-\O_k+0$ in terms of the scaling variable $\th_k$. We
1363: obtain
1364: \bb
1365: \stackrel{n}{X}_k\!\!\!^+\equiv0. \label{Xn}
1366: \ee
1367: The reason of this effect is that all terms connected
1368: with the solutions of the type of $\stackrel{n}{x}_k\!\!\!^+A_1(T_k)$
1369: have been matched by the shift of the independent
1370: variable $\th_k$.
1371: \par
1372: To obtain the formula which connects the constants
1373: $\stackrel{n}{Y}_k\!\!\!^+$ and $\stackrel{m}{y}_k\!\!\!^+$
1374: it is necessary to note that solutions $A_2(T_k)$ is matching
1375: with the solutions $w_2(\th_k)$. Thus the following formula
1376: is valid
1377: \bb
1378: \stackrel{n}{Y}_k\!\!\!^+=\stackrel{m}{y}_k\!\!\!^+,\quad
1379: n=m+7,\quad \stackrel{n}{Y}_k\!\!\!^+=0,\,n=0,1,\dots,6;
1380: \label{Yn}
1381: \ee
1382: The value of the coefficient $\stackrel{7}{Y}_k\!\!\!^+$ is determined
1383: by matching with the asymptotics of $\stackrel{0}{A}(T_k)$
1384: \bb
1385: \stackrel{7}{Y}_k\!\!\!^+=56 g_3(k). \label{Y6}
1386: \ee
1387: \par
1388: Let us represent the expansions of two corrections of the (\ref{weia})
1389: as $\th_k\to\infty$:
1390: \bb
1391: \stackrel{1}{W}(\th_k) = \stackrel{1}{W}_c(\th_k),\label{e1wasy}
1392: \ee
1393: where
1394: $$
1395: \stackrel{1}{W}_c(\th_k) = {{\tau_0 U_*^3}\over 5}\th_k^2 + i{{\tau_0
1396: U_*}\over 5}
1397: \th_k - {\tau_0\over{30U_*}} + i{{3\tau_0}\over{5U_*^3}}\th_k^{-1} +
1398: O(\th_k^{-2})
1399: $$
1400: These formulas explicitly contain the bifurcation point
1401: $\tau_0$. The terms of the asymptotics begin only to depend
1402: on the poles $\O_k,\ k=2,\dots$ of the Weierstrass
1403: $\wp$-function with the second correction term
1404: $\stackrel{2}{W}(\th_1)$. This correction as $\th_k\to\infty$
1405: is represented in the form
1406: \bb
1407: \stackrel{2}{W}(\th_k) =\stackrel{2}{W}_c(\th_k) \label{e2wasy}
1408: \ee
1409: where
1410: $$
1411: \stackrel{2}{W}_c(\th_k)= {{P_k U_*^3}\over 5}\th_1^2 + i{{P_k
1412: U_*}\over 5} \th_k - {P_k\over{30U_*}} +
1413: i{{3P_k}\over{5U_*^3}}\th_k^{-1} + O(\th_k^{-2}),
1414: $$
1415: $$
1416: P_k=\sum_{j=1}^k\O_j, \quad \th_\k \to \infty.
1417: $$
1418: \par
1419: The formulas (\ref{thk}), (\ref{weia}), (\ref{Wn})--
1420: (\ref{Y6}) determine the structure of the formal solution in
1421: the neighborhood of the point $\th_k=+\infty$. It is easy to
1422: see that for $\forall k$ the formal solution (\ref{weia}) is
1423: valid as $\th_k\ve^{1/6}\ll 1$.
1424: \par
1425: Now we know the form of the solution as $\th_k\to\infty$
1426: and two linear independent solutions of the equation in variations
1427: and we are able to write the representation of the correction
1428: terms in the formal solution (\ref{weia}) as $\th_k\to-\infty$.
1429: \bb
1430: \stackrel{n}{W}(\th_k)=\stackrel{n}{W}_c(\th_k) \,+\,
1431: \stackrel{n}{X}_k\!\!\!^-\,w_1(\th_k) \,+\,
1432: \stackrel{n}{Y}_k\!\!\!^-\,w_2(\th_k). \label{Wn-} \ee
1433: Where the functions $\stackrel{n}{W}_c(\th_k)$ have the similar
1434: form as in the formula (\ref{Wn}).
1435: The constants $\stackrel{n}{X}_k\!\!\!^-$ and $\stackrel{n}{Y}_k\!\!\!^-$
1436: are
1437: \bb
1438: \stackrel{n}{X}_k\!\!\!^-=\stackrel{n}{\D}_k\!\!\!^X,\quad
1439: \stackrel{n}{\D}_k\!\!\!^X=-\res_{r=\infty} \bigg[\,{1\over
1440: r}\,\int_{-r}^{r}\big(\stackrel{n}{F}w_2^*+
1441: \stackrel{n}{F}\!\!^*w_2\big){d\th_k\over{\cal W}}\bigg],
1442: \label{Xn+}
1443: \ee
1444: \bb
1445: \stackrel{n}{Y}_k\!\!\!^-=\stackrel{n}{Y}_k\!\!\!^+
1446: +\stackrel{n}{\D}_k\!\!\!^Y,\quad \stackrel{n}{\D}_k\!\!\!^Y=-
1447: \res_{r=\infty}\bigg[\,{1\over
1448: r}\,\int_{-r}^{r}\big(\stackrel{n}{F}w_1^*+
1449: \stackrel{n}{F}\!\!^*w_1\big){d\th_k\over{\cal W}}\bigg].
1450: \label{Yn-}
1451: \ee
1452: It is important to note that
1453: \bb
1454: \stackrel{n}{Y}_k\!\!\!^-=0,
1455: \quad{\hbox{as}}\quad n=1,\dots,6; \label{Y1-5} \ee \bb
1456: \stackrel{7}{Y}_k\!\!\!^-=\stackrel{7}{Y}_k\!\!\!^+ +{\pi \over
1457: 2}. \label{Y0-}
1458: \ee
1459:
1460: \begin{demo} The formula (\ref{Y0-}) in the case of the
1461: second order differential equation was obtained in the work
1462: \cite{Dim-Hab}.
1463: \end{demo}
1464: \par
1465: The domain where the formal series
1466: (\ref{weia}) as $\th_k\to-\infty$ is valid
1467: is determined as well as
1468: $\th_k\to\infty$. We obtain that for any value of $k$ this
1469: domain is determined by $-\th_k\ve^{1/6} \ll 1$ as well as at
1470: large values $-\th_k$.
1471:
1472: \subsubsection{The intermediate expansion, the internal
1473: interval --- interval of the Weierstrass $\wp$-function}
1474: \par
1475: In this section we study the formal asymptotic solution in the
1476: domain between $k$-th and $k+1$-th poles of the Weierstrass
1477: function. The solution in this domain has the asymptotic
1478: structure which is similar to the asymptotic structure in
1479: the initial intermediate interval. The difference is: the
1480: asymptotics on the right side of the investigated interval is
1481: matching with the asymptotics (\ref{weia}) where the
1482: asymptotic sequence is $\{\ve^{n/30}\}$. But the formal
1483: asymptotic solution has been constructed in the intermediate
1484: domain on the initial interval and has been matched with the
1485: asymptotic sequence $\{\ve^{n/5}\}$.
1486: \par
1487: In this intermediate domain we use the slow variable
1488: $T_k=\th_k\ve^{1/6}$ and seek the solution of the equation
1489: (\ref{sh}) in the form
1490:
1491: \bb
1492: U=U_* + \ve^{2/6} A_k(T_k,\ve) + i \ve^{3/6}B_k(T_k,\ve).
1493: \label{A-exp}
1494: \ee
1495: \par
1496: The equation (\ref{sh}) becomes
1497: \begin{eqnarray}
1498: A_k'-2U_*^2 B_k=-\ve^{1/3}2U_*A_kB_k-\ve^{2/3}A_k^2B_k-\ve
1499: B_k^3+\nonumber\\
1500: \bigg(\ve^{4/5}\t_0-\ve^{4/5}\sum_{n=1}^{\infty}\ve^{(2n-1)/5}
1501: \stackrel{n}{\th}_0
1502: +\nonumber\\
1503: \ve^{5/6}\big(T_k-\sum_{j=1}^k\O_j\big)\, -\,
1504: \ve^{5/6}\sum_{n=1}^\infty\ve^{n/30}\sum_{j=1}^k
1505: \stackrel{n}{x}_k\!\!\!^+ \bigg)B_k,
1506: \nonumber\\
1507: B_k'+3U_* A_k^2=-\ve^{1/3}(A_k^3-B_k^2U_*)-\ve^{2/3}A_kB_k^2+
1508: \nonumber\\
1509: \bigg(\ve^{2/15}\t_0+
1510: \ve^{2/15}\sum_{n=1}^\infty\ve^{(2n-1)/5}\stackrel{n}{\th}_0\,+\,
1511: \nonumber\\
1512: \ve^{1/6}\big(T_k-\sum_{j=1}^k\O_j\big)\, -\,
1513: \ve^{1/6}\sum_{n=1}^\infty\ve^{n/30}\sum_{j=1}^k\stackrel{n}{x}_k\!\!\!^+
1514: \bigg)\big(U_*+\ve^{1/3}A_k\big),\quad \ve\to0.
1515: \label{3ineq-k}
1516: \end{eqnarray}
1517: \par
1518: The asymptotics (\ref{weia}) as $\th_{k}\to-\infty$
1519: of the function $U$ has been written in terms of the
1520: new variable $T_k$ gives us the asymptotics as $T_k\to-0$ of the
1521: functions $A$ and $B$
1522: \begin{eqnarray}
1523: A_k=\bigg({-2\over T_k^2}+\stackrel{2}{Y}\!^-_{k-1}
1524: T_k^4+\dots\bigg)+
1525: \ve^{2/15}\bigg({\t_0U_*^3\over5}T_k^2+\dots\bigg)+\nonumber
1526: \\
1527: \ve^{1/6}\bigg({P_kU_*^3\over5}T_k^2+{-1\over6}T_k^3+
1528: \dots\bigg)+\dots,\nonumber\\
1529: B_k=\bigg({-2\over U_*^2T_k^3}+\stackrel{2}{Y}\!^-_{k-1}
1530: T_k^3+\dots\bigg)+ \ve^{2/15}\bigg({\t_0U_*\over5} T_k+\dots
1531: \bigg)+\dots.\label{AB-as-left-k}
1532: \end{eqnarray}
1533: The dots in these formulas mean the terms of the small order
1534: with respect to $T_k$ in the brackets and with respect to
1535: $\ve$ out of them.
1536: $$
1537: P_k=\sum_{j=1}^k\O_j.
1538: $$
1539: \par
1540: The system of the equations (\ref{3ineq-k}) and the asymptotics
1541: (\ref{AB-as-left-k}) contain terms of the different orders with
1542: respect on $\ve$ so we seek the formal asymptotic solution
1543: in the form
1544:
1545: \begin{eqnarray}
1546: A_k(T_k,\ve) = \sum_{n=0}^\infty\ve^{n/30}\stackrel{n}{A}_k(T_k),
1547: \nonumber\\
1548: B_k(T_k,\ve) =
1549: \sum_{n=0}^\infty\ve^{n/30}\stackrel{n}{B}_k(T_k),\quad
1550: \ve\to0.
1551: \label{AB-anzatz-k}
1552: \end{eqnarray}
1553: \par
1554: Using the anzats for the functions $A_k$ and $B_k$ we obtain
1555: the recurrent sequence of the equations for
1556: $\stackrel{n}{A}_k(T_k),\ \stackrel{n}{B}_k(T_k)$. As it was
1557: done above it is convenient for us to investigate the second
1558: order differential equation for $\stackrel{n}{A}_k(T_k)$
1559: instead of the system of the first order for
1560: $\stackrel{n}{A}_k(T_k),\,\stackrel{n}{B}_k(T_k)$. The
1561: equation for the leader term of asymptotics
1562: (\ref{AB-anzatz-k}) is
1563:
1564: \bb
1565: \stackrel{0}{A}_k\!\!^{\prime\prime} + 3\stackrel{0}{A}_k\!^2 = 0.
1566: \label{glp-k}
1567: \ee
1568: The equations for the next correction terms are
1569:
1570: $$
1571: \stackrel{m}{A}_k\!''+6\stackrel{0}{A}_k\stackrel{m}{A}_k=\stackrel{m}{F}.
1572: $$
1573: The expressions for the right hand terms $\stackrel{n}{F}$
1574: depend on $n$. Here we represent the dependness of the
1575: $\stackrel{n}{F}$ on corrections terms with a small
1576: numbers and on the coefficients $\stackrel{j}{x}_k^+$.
1577: The right hand functions $\stackrel{n}{F}$ contain
1578: terms of the type
1579: \newline
1580: with quadratic dependence on corrections $\stackrel{m}{A}_k$
1581: and $\stackrel{l}{A}_k$ as $n=m+l+10$;
1582: \newline
1583: with cubic dependence on corrections $\stackrel{m}{A}_k$,
1584: $\stackrel{j}{A}_k$ and $\stackrel{l}{A}_k$ as $n=m+l+j+20$;
1585: \newline
1586: with linear dependence on corrections $\stackrel{m}{A}_k$ with
1587: coefficients depending on $\stackrel{j}{x}_k^+$ as $n=m+j+15$;
1588: \newline
1589: with linear dependence on correction $\stackrel{m}{A}_k$ with
1590: the linear coefficient with respect to
1591: $(T_k+\sum_{j=1}^k\O_k)$ as $n=m+15$;
1592: \newline
1593: with linear dependence only on $\stackrel{j}{x}\!^+_k$ as
1594: $n=j+5$.
1595: \par
1596: In particular,
1597: \begin{eqnarray*}
1598: \stackrel{m}{F}\equiv0,\,\,m=1,2,3,6,7;\quad
1599: \quad\stackrel{4}{F}=-2U_*^2\t_0;\quad\stackrel{5}{F}=-2U_*^2T_k+P_k;\\
1600: \stackrel{8}{F}=-3\stackrel{4}{A}_k\!\!^2(T_k);\quad\quad\quad
1601: \stackrel{9}{F}=-6\stackrel{4}{A}_k(T_k)\stackrel{5}{A}(T_k);
1602: \\ \stackrel{10}{F}=-\stackrel{1}{\th}_0
1603: U_*-2U_*\bigg({\big(\stackrel{0}{A}_k\!\!'\big)^2\over2U_*}\,+
1604: \,{3\stackrel{0}{A}_k\over2U_*^2}\bigg)
1605: -2U_*^2\bigg(\stackrel{0}{A}_k\!\!\!^3-
1606: {\big(\stackrel{0}{A}_k\!\!'\big)^2\over4U_*^3}\bigg).
1607: \end{eqnarray*}
1608: \par
1609: The asymptotics as $T_k\to-0$ of the coefficients of the
1610: formal series can be obtained from the formulas
1611: (\ref{AB-as-left-k})
1612: \begin{eqnarray*}
1613: \stackrel{0}{A}_k={-2\over T_k^2}+ \stackrel{2}{Y}\!^-_k
1614: T_k^4+\dots;\quad
1615: \stackrel{m}{A}_k\equiv0,\,\,m=1,2,3,6,7;\\
1616: \stackrel{4}{A}_k={\t_0U_*^3\over5}T_k^2+\dots;\quad
1617: \stackrel{5}{A}_k={P_kU_*^3\over5}T_k^2+{-1\over6}T_k^3+\dots;\\
1618: \stackrel{8}{A}_k(T_k)=O(T_k^6);\quad
1619: \stackrel{9}{A}_k=O(T_k^6);\quad \stackrel{10}{A}=O(T_k^{-4}).
1620: \end{eqnarray*}
1621: In these formulas the dots mean the terms of the less order
1622: with respect to $T_k$. In particular, from these formulas we
1623: obtain the domain of suitability for the segment of the series
1624: (\ref{AB-anzatz-k}) as $T_k\to-0$
1625: $$
1626: {\ve^{10/30}\stackrel{10}{A}_k(T_k)\over\stackrel{0}{A}_k(T_k)}\ll1,\,\,\,
1627: {\hbox{or}}\,\,\, -\ve^{-1/6}T_k\gg1.
1628: $$
1629: \par
1630: Let us construct the formal solution in the form
1631: (\ref{AB-anzatz-k}). The seeking solution for the leader term
1632: of the asymptotics equation should have the asymptotic
1633: structure of the type as $T_k\to-0$
1634: $$
1635: \stackrel{0}{A}_k=-{2\over T_k^2}+\stackrel{2}{Y}\!^-_{k-1}
1636: T_k^4+O(T_1^{10}).
1637: $$
1638: Using this formula it is easy to see that the leader term
1639: $\stackrel{0}{A}_k$ of the asymptotics is expressed by the
1640: Weierstrass $\wp$--function
1641:
1642: $$
1643: \stackrel{0}{A}_k(T_k)=-2\wp(T_k;0,g_3(k)),\quad{\hbox{where}}\quad
1644: g_3(k)=\stackrel{2}{Y}\!^-_{k-1}/56.
1645: $$
1646: \par
1647: In a general case the solution $\stackrel{n}{A}(T_k)$ can be
1648: represented in the neighborhood of the point $T_k=0$ in the
1649: form
1650: $$
1651: \stackrel{n}{A}_k(T_k)=\stackrel{n}{A}_k\!\!\!^c(T_k) +
1652: \stackrel{n}{x}_k\!\!\!^- A_1(T_k)+ \stackrel{n}{y}_k\!\!\!^-
1653: A_2(T_k).
1654: $$
1655: The function $\stackrel{n}{A}_k\!\!\!^c(T_k)$ is the partial
1656: solution of the equation for the $n$-th correction term. This solution
1657: does not contain terms of the order of $T_k^{-3}$ and $T_k^4$ in the
1658: asymptotics as $T_k\to 0$.
1659: \par
1660: The coefficients $\stackrel{n}{x}_k\!\!\!^-$ and
1661: $\stackrel{n}{y}_k\!\!\!^-$ are defined by matching of the
1662: asymptotics (\ref{AB-anzatz-k}) as $T_k\to-0$ and the asymptotics
1663: (\ref{weia}) as $\th_k\to\infty$
1664: $$
1665: \stackrel{n}{x}_k\!\!\!^-=\stackrel{m}{X}_k\!\!\!^-,\quad n=6m+23;
1666: $$
1667: $$
1668: \stackrel{n}{y}_k\!\!\!^-=\stackrel{m}{Y}_k\!\!\!^-,\quad n=6m-12.
1669: $$
1670: \par
1671: The period of the function $\stackrel{0}{A}_k(T_k)$ is defined by
1672: the elliptic integral
1673: $$
1674: \O_k=2\int_{A_{min}}^\infty{dy\over\sqrt{4y^3-g_3(k)}}, \quad
1675: A_{min}=\bigg({g_3(k)\over4}\bigg)^{1/3}.
1676: $$
1677: \par
1678: The formal series solution (\ref{AB-anzatz-k}) is not valid
1679: in the neighborhood of the poles. The asymptotics as
1680: $T_k\to0$ of the coefficients of the formal solution is
1681: known. These explicit formulas give us the domain where the
1682: formal solution (\ref{AB-anzatz-k}) is valid. This domain is
1683: $-\ve^{-1/6}T_k\gg1$.
1684: \par
1685: To construct the asymptotics of the coefficients of the
1686: formal series solution in the small neighborhood of the pole
1687: $T_k=-\O_k$ we solve the equations for the corrections and
1688: rewrite their expansions as $T_k\to-\O_k+0$
1689: \par
1690: It is convenient to represent the correction term
1691: $\stackrel{n}{A}_k$ as $T_k\to-\O_k+0$ in the form
1692:
1693: \bb
1694: \stackrel{n}{A}_k(T_k)=\stackrel{n}{A}_c(T_k)+
1695: \stackrel{n}{x}_{k+1}\!\!\!^+ A_1(T_k+\O_k)+
1696: \stackrel{n}{y}_{k+1}\!\!\!^+ A_2(T_k+\O_k).
1697: \label{n-correction-k}
1698: \ee
1699: Where the function $\stackrel{n}{A}_c(T_k)$ is the solution of
1700: the nonhomogeneous equation for the $n$-th correction and its
1701: expansion does not contain terms of the order of
1702: $(T_k+\O_k)^{-3}$ and $(T_k+\O_k)^4$ as $T_k\to-\O_k$.
1703: \par
1704: The constants in the formula for the $n$-th corrections are
1705: evaluated by
1706: $$
1707: \stackrel{n}{y}\!^+_{k+1}=\stackrel{n}{y}_k\!\!\!^-\,+\,
1708: \stackrel{n}{\d}_k\!\!\!^y,\quad \stackrel{n}{\d}_k\!\!\!^y=
1709: \res_{r=0}\bigg[{1\over r} \int_{-r}^{-\O_k+r}
1710: {dz\,\stackrel{n}{F}(z)A_1(z)\over W} \bigg],
1711: $$
1712: $$
1713: \stackrel{n}{x}\!^+_{k+1}= \stackrel{n}{x}_k\!\!\!^-\,+\,
1714: C_k\stackrel{n}{y}\!^+_{k+1}\, +\, \stackrel{n}{\d}_k\!\!\!^x,
1715: \quad \stackrel{n}{\d}_k\!\!\!^x=\res_{r=0}\bigg[{1\over r}
1716: \int_{-r}^{-\O_k+r} {dz\,\stackrel{n}{F}(z)A_2(z)\over W}
1717: \bigg],
1718: $$
1719: where the constant $C_k$ is defined by
1720:
1721: $$
1722: A_2(T_k+\O_k)=C_kA_1(T_k)+A_2(T_k).
1723: $$
1724: \par
1725: To determine the domain of the suitability for the formal
1726: series (\ref{AB-anzatz-k}) as $T_k\to-\O_k+0$ it is important
1727: to know the structure and orders of the singularities of the
1728: correction terms. The first nonzero correction term is
1729: $\stackrel{4}{A}(T_k)$. The representation
1730: (\ref{n-correction-k}) for $n=4$ contains the function
1731: $\stackrel{4}{A}_{k+1}\!\!\!^c(T_k)$ as the partial solution
1732: of the correction equation. This solution has the identical
1733: structure of the asymptotics as $T_1\to\O_1+0$ and as
1734: $T_k\to-0$. But the coefficient
1735: $\stackrel{4}{x}_1\!\!\!^+\not=0$ as $T_1\to\O_1+0$. Namely,
1736:
1737: $$
1738: \stackrel{4}{y}_{k+1}\!\!\!^+=0,
1739: $$
1740: $$
1741: \stackrel{4}{x}_{k+1}\!\!\!^+={28\t_0\over3g_3}\z\big(\O_k/2;0,g_3\big),
1742: $$
1743: where $\z(T_1,g_2,g_3)$ is the Weierstrass $\z$-function.
1744: \par
1745: The values of the coefficients $\stackrel{n}{y}_1\!\!\!^+$ and
1746: $\stackrel{n}{x}_1\!\!\!^+$ for $n>0$ can be also evaluated.
1747: To obtain the domain of the suitability of the formal series
1748: (\ref{AB-anzatz-k}) we note
1749: $$
1750: {\ve^{2/15}\stackrel{4}{A}(T_k)\over\stackrel{0}{A}(T_k)}=
1751: {\ve^{2/15}\over (T_k+\O_k)}\const\ll1.
1752: $$
1753: It is easy to see that the condition
1754: $\ve^{-2/15}(T_k+\O_k)\gg1$ describes the domain where this
1755: solution (\ref{AB-anzatz-k}) is valid as $T_k\to-\O_k$.
1756: \par
1757: Thus in this section we have constructed the formal series
1758: solution in the form (\ref{AB-anzatz-k}) and have shown that
1759: this representation of the solution is valid as
1760: $-\ve^{-1/6}T_k\gg1$ and $\ve^{-2/15}(T_k+\O_k)\gg1$.
1761: \par
1762:
1763: \subsubsection{The discrete dynamical system}
1764:
1765:
1766: \par
1767: The coefficients which determine the intermediate and the
1768: second internal solution are changed by passage from the $k$
1769: to $k+1$. Let us investigate the discrete dynamical system for
1770: the coefficients $\stackrel{n}{X}_k\!\!\!^\pm$,
1771: $\stackrel{n}{Y}_k\!\!\!^\pm$ and $\th_k$,
1772: $\stackrel{n}{x}_k\!\!\!^-$, $\stackrel{n}{y}_k\!\!\!^\pm$.
1773: \par
1774: The connection formulas for these coefficients are
1775: $$
1776: \stackrel{n}{y}\!^+_{k+1}=
1777: \stackrel{m}{Y}\!^-_k\,+\,\stackrel{n}{\delta}\!^y_k,
1778: $$
1779: where $n=6m-12$,
1780: $$
1781: \stackrel{m}{Y}\!^-_{k+1}=\stackrel{n}{y}_k\!\!\!^+\,+\,
1782: \stackrel{m}{\Delta}\!^Y_{k+1}.
1783: $$
1784: Where $\stackrel{m}{Y}_k\equiv0$ as $n=0,\dots,6$. In the
1785: case of $m=7$ we determine the parameter $g_3(k)$ of the
1786: Weierstrass $\wp$--function for the leader term of the
1787: intermediate expansion
1788: $$
1789: g_3(k)={1\over56}\big(a_4+ {\pi\over2} (k-1)\big).
1790: $$
1791: The connection formula is defined by explicit way without
1792: investigating of the next corrections because
1793: at first
1794: $$
1795: \stackrel{0}{\delta}_k\!\!\!^y=0.
1796: $$
1797: and secondly
1798: $$
1799: \stackrel{7}{\Delta}\!^Y_{k+1}={\pi\over2}.
1800: $$
1801: \begin{demo} This connection formula for $g_3(k)$ in the case
1802: of second order ordinary differential equation was found in
1803: the work \cite{Dim-Hab} without investigation of high
1804: correction terms in internal domains and discrete dynamical
1805: system.
1806: \end{demo}
1807: \par
1808: It is easy to see that this system for
1809: $\stackrel{m}{Y}\!^-_{k+1}$ and
1810: $\stackrel{n}{y}_{k+1}\!\!\!^+$ is not closed. It is
1811: connected with the fact that the terms
1812: $\stackrel{m}{\Delta}\!^Y_{k+1}$ and
1813: $\stackrel{n}{\delta}\!^y_{k+1}$ contain the terms
1814: $\stackrel{l}{X}_k\!\!\!^-$ and $\stackrel{j}{x}_k\!\!\!^\pm$.
1815: So to obtain the closed system for the coefficients two above
1816: relations have to supplement by equations for
1817: $\stackrel{l}{X}_k\!\!\!^-$ and $\stackrel{j}{x}_k\!\!\!^\pm$.
1818: As result we obtain
1819: $$
1820: \stackrel{n}{x}_k\!\!\!^+=C_k\stackrel{n}{y}_k\!\!\!^+\,+\,
1821: \stackrel{n}{\delta}{}^x_{k-1}
1822: \,+\,\stackrel{n}{\Delta}{}^X_{k-1}.
1823: $$
1824: \par
1825: This discrete dynamical system allows us to evaluate the
1826: phase variable in the leader term of the second internal
1827: expansion
1828: $$
1829: \th_{k+1}=\th_k+\ve^{-1/6}\bigg(\O_{k+1}+
1830: {1\over4}\sum_{n=1}^\infty
1831: \ve^{n/30}\stackrel{n}{x}_{k+1}^+\bigg),\quad \ve\to0.
1832: $$
1833: \par
1834: Thus, to obtain the phase variable to within $o(1)$ in the
1835: neighborhood of the $(k+1)$-th pole it is necessary to
1836: obtain the term $\stackrel{n}{x}_{k+1}\!\!\!^+$ as $n\le5$ and
1837: $\stackrel{m}{Y}_k\!\!\!^-$ as $m\le3$.
1838: \par
1839: Let us consider these formulas as $n\le5$ and $m\le3$
1840: $$
1841: \stackrel{m}{\D}_{k}\!\!\!^Y\equiv0,\quad m\not=7; $$ $$
1842: \stackrel{n}{\d}_k\!\!\!^y\equiv0,\quad 0<n<5; $$ $$
1843: \stackrel{5}{\d}_k\!\!\!^y=2\z(\O_k/2,0,g_3(k)). $$
1844: We also have that
1845: $\stackrel{l}{\D}_k\!\!\!^X\equiv0,$ as $l<1$.
1846: \par
1847: As result the formulas for $\stackrel{n}{x}_k\!\!\!^+$
1848: can be written in the form
1849: $$
1850: \stackrel{n}{y}_k\!\!\!^+\equiv0\quad
1851: \stackrel{n}{x}_k\!\!\!^+\equiv0,\quad n=1,2,3,$$ $$
1852: \stackrel{4}{y}_k\!\!\!^+\equiv0,\quad
1853: \stackrel{4}{x}_k\!\!\!^+=\stackrel{4}{\d}_k\!\!\!^x,$$ $$
1854: \stackrel{5}{y}_k\!\!\!^+=\stackrel{5}{y}\!^+_{k-1} \,+\,
1855: \stackrel{5}{\d}\!^y_{k-1},\quad
1856: \stackrel{5}{x}\!^+_k=C_k\stackrel{5}{y}\!^+_k \,+\,
1857: \stackrel{5}{\d}\!^x_k.
1858: $$
1859:
1860:
1861: \subsubsection{The domain of the suitability of the intermediate
1862: expansion}
1863:
1864: \par
1865: The domain where the intermediate asymptotic
1866: expansion is valid for big values of $k$ is defined by inequality
1867: $$
1868: {\ve^{1/6}\stackrel{5}{A}_k\over\stackrel{0}{A}_k}\ll1.$$
1869: \par
1870: The function $\stackrel{5}{A}_k$ contains the term
1871: $\stackrel{5}{y}_k\!\!\!^+ A_1(T_k)$ which is the biggest
1872: with respect to $k$
1873: $$
1874: \stackrel{5}{y}_k\!\!\!^+= \sum_{j=1}^k\stackrel{5}{\d}_j\!\!\!^y
1875: =O(k^{7/6})$$
1876: \par
1877: This formula leads us to the inequality
1878: $$ \ve^{1/6}k^{7/6}\ll1,\quad
1879: k\ll\ve^{-1/7}.
1880: $$
1881: It is easy to determine the order of
1882: $P_k$ as $k\to \infty$
1883: $$
1884: P_k=\sum_{j=1}^k\O_j=O(k^{5/6}),\quad \hbox{ÐÒÉ}\quad k\to\infty.
1885: $$
1886:
1887:
1888:
1889: \subsection{The intermediate expansion for large $k$}
1890:
1891: \par
1892: In this section to take into account the secular terms as
1893: large values of $k$ we modify the intermediate expansion.
1894:
1895: \subsubsection{The domain for moderate large values of $k$}
1896:
1897: \par
1898: As was shown above for the large values of $k$ the correction
1899: terms of the asymptotics contain the secular terms with
1900: respect to $k$. These singularities have the different
1901: structure. At first, the singularities are connected with the
1902: terms $\stackrel{n}{y}_k\!\!\!^+$ as the solutions of the
1903: homogeneous equations. Secondly, the singular terms $P_k$ and
1904: $\sum_{j=1}^k\stackrel{n}{x}_j\!\!\!^+$ are contained in
1905: right-hand sides of the equations (\ref{3ineq-k}). These
1906: singularities lead to lose of the suitability of the
1907: equations for the correction terms equations.
1908: \par
1909: To suppress these singularities we allow the dependness
1910: of the parameter $g_3$ on the small parameter
1911: $$
1912: g_3(k,\ve)=\stackrel{0}{g}_3(k)+\sum_{n=1}^\infty
1913: \ve^{n/30}\stackrel{n}{g}_3(k).
1914: $$
1915: Where
1916: $$
1917: \stackrel{n}{g}_3(k+1)=\stackrel{n}{g}_3(k)+{1\over56}\big(
1918: \stackrel{m}{\D}_k\!\!\!^Y+\stackrel{n}{\d}_k\!\!\!^y\big),\quad m=n+5.
1919: $$
1920: \par
1921: The equations for $\stackrel{n}{x}_k\!\!\!^+$ become
1922: $$
1923: \stackrel{n}{x}_k\!\!\!^+=C_k\stackrel{n}{\d}_k\!\!\!^y \,+\,
1924: \stackrel{l}{\D}\!^X_k,\quad n=l+29.
1925: $$
1926: We can suppress the first kind of the singularities in the
1927: intermediate expansion by this way. To suppress the
1928: singularities of the second type connected with large values
1929: of $P_k$ and sum of the coefficients
1930: $\stackrel{n}{x}_j\!\!\!^+$ we include the following term in
1931: equation for the leader term of the intermediate expansion
1932: $$
1933: \l_k(\ve)=\ve^{1/6}\bigg(P_k+{1\over4}\sum_{n=1}^\infty\ve^{(n-1)/30}
1934: \sum_{j=1}^k\stackrel{n}{x}_j\!\!\!^+\bigg).
1935: $$
1936: Then the leader term equation has the form
1937: $$
1938: \stackrel{0}{A}_k''+3\stackrel{0}{A}_k\!\!\!^2=\l_k,
1939: $$
1940: \par
1941: and the solution is
1942: $$
1943: \stackrel{0}{A}_k(T_k)=-2\wp(T_k,\l_k/2,g_3(k,\ve)).
1944: $$
1945: \par
1946: Including of the term $\l_k(\ve)$ in equation for the leader term
1947: of the asymptotics leads us to change of the asymptotic sequence.
1948: Namely, $$ A(T_k,\ve)=\sum_{n=0}^\infty
1949: \ve^{n/6}\stackrel{5n}{A}(T_k). $$ The terms of the order of
1950: $\ve^{m/30}$ with $m\not= 5l, l\in Z$ are contained in the leader
1951: term of the asymptotics and corrections by introducing of the
1952: $\l_k(\ve)$. To obtain the intermediate expansion with respect to
1953: $\ve^{m/30}$ one should expand this representation with respect
1954: to $\ve$.
1955: \par
1956: The solutions of the corrections equations are defined by
1957: their asymptotics as $T_k\to-0$:
1958: $$
1959: \stackrel{5n}{A}_k(T_k)=\stackrel{5n}{A}_k\!\!\!^c(T_k)+
1960: \stackrel{5n}{x}_k\!\!\!^- A_1(T_k)
1961: $$
1962: The main term of the asymptotics with respect to $T_k$ is the
1963: term of the order of $T_k^4$. The function
1964: $\stackrel{n}{A}_k$ is the partial solution of the
1965: nonhomogeneous equation for the correction term and does not
1966: contain the terms of the order of $T_k^4$ and $T_k^{-3}$ in
1967: the asymptotics as $T_k\to-0$. The function $A_1$ is the
1968: solution of variation equation and is defined by the leader
1969: term
1970: $$
1971: A_1={d\over dT_k}\wp(T_k,\l_k/2,g_3(k,\ve)).
1972: $$
1973: \par
1974: In the neighborhood of the point $T_k=-\O_k+0$ the corrections
1975: can be represented by formulas
1976: $$
1977: \stackrel{n}{A}_k(T_k)=\stackrel{n}{A}_k\!\!\!^c(T_k)+
1978: \stackrel{n}{x}\!^+_{k+1} A_1(T_k) \,+\,
1979: \stackrel{n}{\d}\!^y_k A_2(T_k).
1980: $$
1981: The function $A_2$ is the second solution of the homogeneous
1982: variation equation
1983: $$
1984: A_2={d\over dg_3}\wp(T_k,\l_k/2,g_3),\quad \hbox{ÐÒÉ}\quad
1985: g_3=g_3(k,\ve).
1986: $$
1987: \par
1988: The constructed solution is valid between the poles of the
1989: Weierstrass $\wp$-function $\wp(T_k,\l_k/2,g_3(k,\ve))$ in
1990: the similar domains as in the case of intermediate expansion
1991: for finite values of $k$.
1992:
1993: \subsubsection{The second internal expansion for moderate
1994: large $k$}
1995: \par
1996: To construct the second internal expansion in this case
1997: we should make the changes in the formula for evaluating of
1998: $\stackrel{n}{Y}_k\!\!\!^+$:
1999: $$
2000: \stackrel{n}{Y}_k\!\!\!^+=56\stackrel{m}{g}_3(k), \quad n=6m-12.
2001: $$
2002:
2003:
2004: \subsubsection{The large values of $k$}
2005:
2006: \par
2007: The large values of $k$ lead to grow of the term $\l_k(\ve)$
2008: in the right hand side of the equation for leader term of the
2009: asymptotics. The structure of the formal solution in this
2010: case becomes complicated. Two parameters appear and the
2011: formal serires solution of the original equation use both of
2012: them. These two parameters are $\ve$ as small one and
2013: $\l_k(\ve)$ as large one. To simplify the representation of
2014: the formal series solution in this section we only use the
2015: segments of the formal asymptotic series. We describe the
2016: domain of the suitability of this representation of the
2017: formal series solution. It is convenient to construct the
2018: leader term of the intermediate asymptotics in the form
2019: $$
2020: A_k=\sqrt{\l_k}\big(\stackrel{0}{\cal A}_k \,+\,
2021: \ve^{1/6}\stackrel{5}{\cal A}_k\big) \,+\,
2022: \ve^{1/3}\l_k\big(\stackrel{10}{\cal A}_k \,+\,
2023: \ve^{1/2}\stackrel{15}{\cal A}_k\big) \,+\,
2024: \ve^{2/3}\l_k^{3/2}\stackrel{20}{\cal A}_k,
2025: $$
2026: $$
2027: B_k=\l_k^{3/4}\big(\stackrel{0}{\cal B}_k \,+\,
2028: \ve^{1/6}\stackrel{5}{\cal B}_k\big) \,+\,
2029: \ve^{1/3}\l_k^{5/4}\big(\stackrel{10}{\cal B}_k \,+\,
2030: \ve^{1/6}\stackrel{15}{\cal B}_k\big) \,+\,
2031: $$
2032: $$
2033: \ve^{2/3}\l_k^{7/4}\big(\stackrel{20}{\cal B}_k+
2034: \ve^{1/6}\stackrel{25}{\cal B}_k\big).
2035: $$
2036: The coefficients of the asymptotics contain the independent
2037: variable with the large term $\s_k=\l_k^{1/4}T_k$, namely,
2038: $\stackrel{n}{\cal A}_k=\stackrel{n}{\cal A}_k(\s_k,k,\ve)$
2039: and
2040: $\stackrel{n}{\cal B}_k=\stackrel{n}{\cal B}_k(\s_k,k,\ve)$.
2041: \par
2042: The leader term of the asymptotics satisfies to the equation
2043: \begin{eqnarray*}
2044: \stackrel{0}{\cal A}_k'-2U_*\stackrel{0}{\cal B}_k=0,\\
2045: \stackrel{0}{\cal B}_k'+3U_*\stackrel{0}{\cal A}_k\!\!\!^2= U_*.
2046: \end{eqnarray*}
2047: It is easy to write the equations for the correction terms
2048: of the asymptotics
2049: \begin{eqnarray*}
2050: \stackrel{5}{\cal A}_k'-2U_*\stackrel{5}{\cal B}_k=0,\\
2051: \stackrel{5}{\cal B}_k'+6U_*\stackrel{0}{\cal A}_k\stackrel{5}{\cal A}_k=
2052: U_*T_k/\l_k;
2053: \end{eqnarray*}
2054: \begin{eqnarray*}
2055: \stackrel{10}{\cal A}_k'-2U_*\stackrel{10}{\cal B}_k=
2056: -2U_*\stackrel{0}{\cal A}_k\stackrel{0}{\cal B}_k,\\
2057: \stackrel{10}{\cal B}_k'+6U_*\stackrel{0}{\cal
2058: A}_k\stackrel{10}{\cal A}_k= -3U_*\stackrel{5}{\cal A}_k\!\!\!^2
2059: \,+\, \stackrel{0}{\cal A}_k - \stackrel{0}{\cal A}_k\!\!\!^3
2060: -U_*\stackrel{0}{\cal B}_k\!\!\!^2;
2061: \end{eqnarray*}
2062: \begin{eqnarray*}
2063: \stackrel{15}{\cal A}_k'-2U_*\stackrel{15}{\cal B}_k=
2064: -2U_*\big(\stackrel{5}{\cal A}_k\stackrel{0}{\cal B}_k+
2065: \stackrel{0}{\cal A}_k\stackrel{5}{\cal B}_k\big),\\
2066: \stackrel{15}{\cal B}_k'+6U_*\stackrel{0}{\cal A}_k\stackrel{15}{\cal A}_k=
2067: \big(1-3\stackrel{0}{\cal A}_k\!\!\!^2\big)\stackrel{5}{\cal A}_k
2068: \,+\,
2069: {T_k\over\l_k}\stackrel{0}{\cal A}_k
2070: \,- \,
2071: 2U_*\stackrel{0}{\cal B}_k\stackrel{5}{\cal B}_k
2072: \,-\,
2073: 6U_*\stackrel{5}{\cal A}_k\stackrel{10}{\cal A}_k;
2074: \end{eqnarray*}
2075: \begin{eqnarray*}
2076: \stackrel{20}{\cal A}_k\!'-2U_*\stackrel{20}{\cal B}_k=
2077: -\stackrel{0}{\cal B}_k-\stackrel{0}{\cal A}_k\!\!\!^2
2078: \stackrel{0}{\cal B}_k -
2079: \\
2080: 2U_*\big(\stackrel{10}{\cal A}_k\stackrel{0}{\cal B}_k \,+\,
2081: \stackrel{0}{\cal A}_k\stackrel{10}{\cal B}_k \,+\,
2082: {1\over\sqrt{\l_k}}\stackrel{5}{\cal A}_k\stackrel{5}{\cal
2083: B}_k\big),
2084: \\
2085: \stackrel{20}{\cal B}_k\!'+6U_*\stackrel{0}{\cal
2086: A}_k\stackrel{20}{\cal A}_k= \stackrel{10}{\cal
2087: A}_k-3\stackrel{0}{\cal A}_k\!\!\!^2\stackrel{10}{\cal A}_k-
2088: \stackrel{0}{\cal A}_k\stackrel{0}{\cal B}_k\!\!\!^2-
2089: \\
2090: 2U_*\stackrel{0}{\cal B}_k\stackrel{10}{\cal B}_k \,+\,
2091: {1\over\l_k^{1/2}} \big(-3 \stackrel{0}{\cal A} \stackrel{5}{\cal
2092: A}_k\!\!\!^2 - U_*\stackrel{5}{\cal B}_k\!\!\!^2\big)
2093: \\
2094: \,+\,{1\over\l_k^{3/2}} T_k\stackrel{5}{\cal A}_k
2095: \,-\,
2096: 3U_*\stackrel{10}{\cal A}_k\!\!\!^2- 6U_*\stackrel{5}{\cal
2097: A}_k\stackrel{15}{\cal A}_k.
2098: \end{eqnarray*}
2099: \par
2100: As was shown above it is convenient to investigate the second order
2101: differential equation for the functions $\stackrel{n}{\cal A}_k$.
2102: In particular, we have
2103:
2104: $$
2105: \stackrel{0}{\cal A}_k\!\!\!''+3\stackrel{0}{\cal A}_k\!\!\!^2=1.
2106: $$
2107: The solution of this equation matched with the asymptotics of the internal
2108: formal solution in the neighborhood of the $k-1$-th pole is
2109: $$
2110: \stackrel{0}{\cal A}=-2\wp(\s_k,1/2,\g_3(k,\ve)),
2111: \quad\hbox{where}\quad \g_3(k,\ve)={g_3(k,\ve)\over\l_k^{3/2}}.
2112: $$
2113: \par
2114: The parameter $g_3(k,\ve)$ and corrections of the asymptotics are
2115: obtained as was shown above in the domain of the intermediate
2116: expansion. It is easy to see that unique type of singularities
2117: is connected with the parameter $\l_k$. The intermediate
2118: expansion is valid as
2119: $$
2120: {\ve^{1/3}\l_k\over\sqrt{\l_k}}\ll1,\quad \hbox{or}\quad \l_k\ll\ve^{-2/3}.
2121: $$
2122: \par
2123: The constructed sequence of the intermediate and the second internal
2124: expansions is valid in the domain as
2125: $$
2126: (t_*-t)\ll\ve^{1/6}.
2127: $$
2128:
2129: \section{Fast oscillating asymptotic solution}
2130:
2131: In this section we construct a special fast oscillating
2132: asymptotic solution of the equation (\ref{sh}) by
2133: Krylov-Bogolyubov method \cite{K-B}, \cite{Kuz}. This
2134: constructed solution is valid as $(t_*-t)\ve^{-2/3} \gg 1$.
2135: This special solution degenerates at the point $t=t_*$. In the
2136: neighborhood of this point this special solution loses the
2137: fast oscillating structure and matchs with the asymptotics
2138: from the previous section.
2139:
2140:
2141: \subsection{The family of the fast oscillating solutions}
2142:
2143: Here we construct the family of the fast oscillating solutions
2144: which depends on two parameters.
2145: \par
2146: The fast oscillating solution of the equation
2147: (\ref{sh}) is seeking in the form
2148: \bb
2149: U(t,\ve) = \stackrel{0}{U}(t_1,t,\ve) + \ve
2150: \stackrel{1}{U}(t_1,t,\ve) + \ve^2 \stackrel{2}{U}(t_1,t,\ve) +
2151: \dots, \label{as4}
2152: \ee
2153: where $t_1=S(t)/\ve+\phi(t)$ is the fast variable, the functions
2154: $S(t)$ and $\phi(t)$ are unknown functions.
2155: \par
2156: The problem on constructing the fast oscillating solution for
2157: the second order differential equation was investigated by
2158: many authors. In this section we construct the solution in
2159: the manner of F.J.Bourland and R.Haberman \cite{Bour-Hab2}.
2160: They have investigated the second order equation so the
2161: direct reference for the equation (\ref{sh}) is not valid. In
2162: this section we represent some evaluations which lead us to
2163: the elegant formulas from the work \cite{Bour-Hab2}.
2164: \par
2165: Let us substitute (\ref{as4}) in equation (\ref{sh}). As a result
2166: we obtain the equations for the leader term of the asymptotics
2167: (\ref{as4}) and the first correction term. \bb
2168: iS^{\prime}\pt_{t_1} \stackrel{0}{U} +
2169: |\stackrel{0}{U}|^2\stackrel{0}{U} - t \stackrel{0}{U} = 1.
2170: \label{u0} \ee \bb iS^{\prime}\pt_{t_1} \stackrel{1}{U} +
2171: \left(2|\stackrel{0}{U}|^2-t\right) \stackrel{1}{U} +
2172: \stackrel{0}{U}\!^2\stackrel{1}{U^*} = -i\pt_t \stackrel{0}{U}
2173: -i\phi'\stackrel{0}{U}. \label{u1} \ee
2174: \begin{eqnarray}
2175: iS^{\prime}\pt_{t_1} \stackrel{2}{U} +
2176: \left(2|\stackrel{0}{U}|^2-t\right) \stackrel{2}{U} +
2177: \stackrel{0}{U}\!^2\stackrel{2}{U^*} = \nonumber
2178: \\
2179: -i\pt_t
2180: \stackrel{1}{U}-i\phi'\pt_{t_1}\stackrel{1}{U}-
2181: 2|\stackrel{1}{U}|^2\stackrel{0}{U}-
2182: (\stackrel{1}{U})^2\stackrel{0}{U}\!^*. \label{u2}
2183: \end{eqnarray}
2184: \par
2185: The time variable $t$ is the independent variable in these
2186: equations. The equation (\ref{u0}) has the first integral
2187: with respect to "fast" variable $t_1$
2188:
2189: \bb
2190: {1\over 2}|\stackrel{0}{U}|^4 -t |\stackrel{0}{U}|^2 -
2191: (\stackrel{0}{U} + \stackrel{0}{U^*}) = E(t), \label{int}
2192: \ee
2193:
2194: where $E(t)$ is a "constant" of integration.
2195:
2196: \par
2197: This expression for the first integral can be considered as
2198: the equation for the function $\stackrel{0}{U}$. Let us
2199: express $\stackrel{0}{U}$ through the complex conjugate
2200: function and as a result we obtain the equation only for
2201: $\stackrel{0}{U}$.
2202:
2203: $$
2204: iS^{\prime}\pt_{t_1} \stackrel{0}{U} = \pm
2205: \sqrt{2\stackrel{0}{U}\!^3 + (2E(t)+t^2) \stackrel{0}{U}\!^2 +
2206: 2t\stackrel{0}{U}+1}.
2207: $$
2208: This equation can be easy integrated
2209:
2210: \bb
2211: iS'\int_{u_0}^{\stackrel{0}{U}}
2212: {dy\over\pm\sqrt{2y^3+(2E+t^2)y^2+2ty+1}}= t_1+S_0.
2213: \label{u0s}
2214: \ee
2215: Where $u_0$ and $S_0$ are constants and we integrate on the
2216: curve $\G(t)$ on the complex plane $y$. This curve is defined
2217: by
2218:
2219: $$
2220: {1\over 2}|y|^4 -t |y|^2 - (y + y^*) = E(t).
2221: $$
2222: The sign $+$ or $-$ fixed the sheet of a Riemann surface
2223: where the initial point $u_0$ is situated. We choose the "$+$"
2224: sign.
2225:
2226: \par
2227: The motion defined by (\ref{u0s}) is periodic with respect to
2228: fast variable $t_1$. By integrating of the formula (\ref{u0s})
2229: over the period we obtain:
2230: \bb
2231: iS'\int_{\G(t)}{dy\over\sqrt{2y^3+(2E+t^2)y^2+2ty+1}}=T.
2232: \label{period}
2233: \ee
2234: The constant $T$ is the period of the function $\stackrel{0}{U}$
2235: with respect to fast variable $t_1$. This formula
2236: is the differential equation for the function $S(t)$.
2237:
2238: \par
2239: It is convenient to represent the leader term
2240: $\stackrel{0}{U}(t_1,t)$ as the sum of the real and the
2241: imaginary parts
2242: $$
2243: \stackrel{0}{U}=\stackrel{0}{U}_R+i\stackrel{0}{U}_I.
2244: $$
2245: By shift of phase variable $t_1$ we can obtain that
2246: $\stackrel{0}{U}_R(t_1,t)$ is even function with respect
2247: to $t_1$ and $\stackrel{0}{U}_I(t_1,t)$ is odd one.
2248: This property of the functions $\stackrel{0}{U}_I(t_1,t)$
2249: and $\stackrel{0}{U}_R(t_1,t)$ is convenient
2250: to evaluate the integrals with respect to fast
2251: variable.
2252:
2253: \par
2254: To determine the leader term of the asymptotics (\ref{as4})
2255: it is necessary to obtain an equation for $E(t)$. This
2256: equation appears from a condition of the boundedness of the
2257: first correction term $\stackrel{1}{U}$ of the asymptotics.
2258: \par
2259: The general solution of the equation for the first correction
2260: can be represented by two linear independent solutions of the
2261: homogeneous equation
2262:
2263: $$ u_1(t_1,t)=\pt_{t_1}\stackrel{0}{U},\quad \hbox{and}\quad
2264: u_2(t_1,t)={\pt_ES\over S'}t_1 u_1 +\pt_E \stackrel{0}{U}. $$ In
2265: these formulas the $u_1$ is periodic bounded function, but $u_2$
2266: grows linearly with respect to $t_1$. The Wronskian of these
2267: solutions is
2268:
2269: $$
2270: W=u_1u_2^*-u_1^*u_2={1\over iS'}.
2271: $$
2272: It can be evaluated by explicit formulas
2273: $$
2274: W={1\over iS'} \big(-(|\stackrel{0}{U}|^2-t)\stackrel{0}{U}+1\big)
2275: \pt_E\stackrel{0}{U}\!^* + {1\over iS'}
2276: \big(-(|\stackrel{0}{U}|^2-t)\stackrel{0}{U}\!^*+1\big)
2277: \pt_E\stackrel{0}{U}=
2278: $$
2279: $$
2280: = {1\over iS'}\pt_E\big(-{1\over2}
2281: |\stackrel{0}{U}|^4+t|\stackrel{0}{U}|^2+\stackrel{0}{U}
2282: +\stackrel{0}{U}\!^*\big)={-1\over iS'}.
2283: $$
2284: \par
2285: To obtain necessary condition of the boundedness of the
2286: first correction terms of the asymptotics (\ref{as4}) we
2287: multiply the equation (\ref{u1}) to
2288: $\pt_{t_1}\stackrel{0}{U}\!^*$ and integrate it with respect
2289: to the variable $t_1$ over the period. Then we integrate the
2290: first term by parts
2291:
2292: $$
2293: iS'\int_{0}^Td t_1 \pt_{t_1}\stackrel{1}{U}\pt_{t_1}
2294: \stackrel{0}{U}\!^*=-\int_0^T dt_1\bigg(2(|\stackrel{0}{U}|^2-t)
2295: \pt_{t_1}\stackrel{0}{U}\!^*+\big(\stackrel{0}{U}\!^*\big)^2
2296: \pt_{t_1}\stackrel{0}{U}\bigg)\stackrel{1}{U}.
2297: $$
2298: As a result we obtain:
2299:
2300: $$
2301: \int_0^Tdt_1\bigg[\stackrel{0}{U}\!^2\stackrel{1}{U}\!^*
2302: \pt_{t_1}\stackrel{0}{U}\!^*-\big(\stackrel{0}{U}\!^*\big)^2
2303: \pt_{t_1}\stackrel{0}{U}\stackrel{1}{U}\bigg]=
2304: $$
2305: $$
2306: -i\int_0^Tdt_1\pt_t\stackrel{0}{U}\pt_{t_1}\stackrel{0}{U}\!^*
2307: -i\phi'\int_0^Tdt_1
2308: \pt_{t_1}\stackrel{0}{U}\pt_{t_1}\stackrel{0}{U}\!^*.
2309: $$
2310: \par
2311: Let us consider the equation for the complex conjugate
2312: function
2313:
2314: $$
2315: -iS^{\prime}\pt_{t_1} \stackrel{1}{U}\!^* +
2316: \left(2|\stackrel{0}{U}|^2-t\right) \stackrel{1}{U}\!^* +
2317: \big(\stackrel{0}{U}\!^*\big)\!^2\stackrel{1}{U}= i\pt_t
2318: \stackrel{0}{U}\!^*+i\phi\pt_{t_1}\stackrel{0}{U}\!^*.
2319: $$
2320: Multiply this equation to $\pt_{t_1}\stackrel{0}{U}$ and
2321: integrate it with respect to $t_1$ over the period. Then we
2322: obtain
2323: $$
2324: \int_0^Tdt_1\bigg[\big(\stackrel{0}{U}\!^*\big)^2\stackrel{1}{U}
2325: \pt_{t_1}\stackrel{0}{U}-\stackrel{0}{U}\!^2
2326: \pt_{t_1}\stackrel{0}{U}\!^*\stackrel{1}{U}\!^*\bigg]=
2327: $$
2328: $$
2329: i\int_0^Tdt_1\pt_t\stackrel{0}{U}\!^*\pt_{t_1}\stackrel{0}{U} +
2330: i\phi'\int_0^Tdt_1
2331: \pt_{t_1}\stackrel{0}{U}\pt_{t_1}\stackrel{0}{U}\!^*.
2332: $$
2333: Combining obtained expressions we get
2334:
2335: $$
2336: -i\int_0^Tdt_1 \pt_t\stackrel{0}{U}\pt_{t_1}\stackrel{0}{U}\!^*+
2337: i\int_0^Tdt_1 \pt_t\stackrel{0}{U}\!^*\pt_{t_1}\stackrel{0}{U}=0.
2338: $$
2339: After integrating by parts we get
2340:
2341: $$
2342: i\int_0^Tdt_1 \stackrel{0}{U}\!^*\pt_{t_1}\pt_t\stackrel{0}{U}+
2343: i\int_0^Tdt_1\pt_t\stackrel{0}{U}\!^*\pt_{t_1}\stackrel{0}{U}=0,
2344: $$
2345: or
2346: $$
2347: i\pt_t\int_0^Tdt_1\stackrel{0}{U}\!^*\pt_{t_1}\stackrel{0}{U}=0.
2348: $$
2349: or
2350: \bb
2351: I\equiv i\int_{\G(t)}u^*du=\s=\const. \label{inv}
2352: \ee
2353: The integral in this formula geometrically means the doubled
2354: square of the domain bounded by the curve $\G(t)$. The
2355: necessary condition of the boundedness of the first
2356: correction term is invariance of this square. It is the
2357: adiabatic invariant.
2358: \par
2359: The solution of the equation for the first correction term
2360: can be written in the form
2361:
2362: \bb
2363: \stackrel{1}{U}(t_1,t)=c_1\pt_{t_1}\stackrel{0}{U}+
2364: {-i\phi'\over\pt_E S}\pt_E\stackrel{0}{U}+\stackrel{1}{R}(t_1,t).
2365: \label{u1-form}
2366: \ee
2367: \par
2368: Where
2369: $$
2370: \stackrel{1}{R}(t_1,t)=u_1(t_1,t)\int_0^{t_1}
2371: dt'\big(\pt_t\stackrel{0}{U}(t',t)u_2^*(t',t)+
2372: \pt_t\stackrel{0}{U}\!^*(t',t)u_2(t',t)\big) \,-\,
2373: $$
2374: $$
2375: \,-\, u_2(t_1,t)\int_0^{t_1}
2376: dt'\big(\pt_t\stackrel{0}{U}(t',t)u_1^*(t',t)+
2377: \pt_t\stackrel{0}{U}\!^*(t',t)u_1(t',t)\big).
2378: $$
2379: To evaluate the second correction we use the property of
2380: even and odd of the real and imaginary parts of the first
2381: correction term. These properties of the first and the second
2382: terms in the formula (\ref{u1-form}) can be obtained from analysis
2383: of the leader term of the asymptotics. The last term of this sum
2384: contains the even function $\Re(\stackrel{1}{R})$ and
2385: the odd function $\Im(\stackrel{1}{R})$ with respect to variable
2386: $t_1$.
2387: \par
2388: Constructing of the periodical second correction of the
2389: asymptotics leads us to determining of the function $\phi(t)$.
2390: The necessary condition of this property for the second
2391: correction is $$ \int_0^Tdt_1\bigg(\stackrel{2}{F}
2392: \pt_{t_1}\stackrel{0}{U}\!^* +\stackrel{2}{F}\!^*
2393: \pt_{t_1}\stackrel{0}{U}\bigg)=0. $$ Where $\stackrel{2}{F}$
2394: is the right-hand side of the equation (\ref{u2}).
2395: \par
2396: Let us separately evaluate the integral depending on square
2397: of the term $\stackrel{1}{U}$
2398:
2399: $$
2400: J=\int_0^Tdt_1\big[-(2|\stackrel{1}{U}|^2\stackrel{0}{U}+
2401: \stackrel{1}{U}\!^2\stackrel{0}{U}\!^*) \pt_{t_1}\stackrel{0}{U}\!^*
2402: \,-\, (2|\stackrel{1}{U}|^2\stackrel{0}{U}\!^*+
2403: \big(\stackrel{1}{U}\!^*\big)^2\stackrel{0}{U})
2404: \pt_{t_1}\stackrel{0}{U} \big].
2405: $$
2406: Integrating by parts and substituting of the expressions
2407: for $\pt_{t_1}\stackrel{1}{U}$ leads us to the following formula
2408: $$
2409: J=\int_0^Tdt_1\big[\big(-i\pt_t\stackrel{0}{U}
2410: -i\phi'\pt_{t_1}\stackrel{0}{U}\big) \pt_{t_1}\stackrel{1}{U}\!^*
2411: +\big(-i\pt_t\stackrel{0}{U}\!^*
2412: +i\phi'\pt_{t_1}\stackrel{0}{U}\!^*\big) \pt_{t_1}\stackrel{1}{U}
2413: \big]
2414: $$
2415: Then we obtain
2416: $$
2417: \pt_t\int_0^Tdt_1\big[i\stackrel{1}{U}\!^*\pt_{t_1}\stackrel{0}{U}
2418: \,-\,i\stackrel{1}{U}\pt_{t_1}\stackrel{0}{U}\!^*\big].
2419: $$
2420: Substituting of $\stackrel{1}{U}$ in this formula and
2421: using of even properties gives us
2422: $$
2423: \pt_t\int_0^Tdt_1\,{\phi'\over\pt_E
2424: S}\big(\pt_E\stackrel{0}{U}\pt_{t_1}\stackrel{0}{U}\!^*\big)=0
2425: $$
2426: or \bb {\phi'\over \pt_E S}\pt_E I \equiv\phi_1=\const.
2427: \label{phi'} \ee
2428: This elegant form of the equation for the phase shift was
2429: obtained by F.J.Bo\-ur\-land É R.Haberman \cite{Bour-Hab2} for
2430: the solutions of the second order differential equation.
2431: \par
2432: The formulas (\ref{u0s}), (\ref{period}), (\ref{inv}) and
2433: (\ref{phi'}) determine the family of the fast oscillating
2434: asymptotic solutions of the equation (\ref{sh}). The
2435: solutions of this family contain three parameters. The
2436: constant $\s$ determines the trajectory on the phase plane.
2437: The constant $\phi_1$ in the formula (\ref{phi'}) and initial
2438: condition $\phi|_{t=\const}=\phi_0$ determine the phase
2439: shift. The periodic with respect to $t_1$ leader term of the
2440: asymptotics depends on three constants. Specificity of this
2441: fact is explained by the large values of the time variable.
2442: So to determine the phase shift correctly we should take into
2443: account one condition for the first correction term of the
2444: constructed asymptotics. The period $T$ is arbitrary nonzero
2445: constant.
2446:
2447:
2448: \subsection{The confluent asymptotic solution}
2449:
2450: \par
2451: In this section we choose the values of the parameters when
2452: the constructing fast oscillating solution degenerates at the
2453: moment $t=t_*$. The algebraic solution constructed above
2454: degenerates at the moment $t=t_*$ in the point $U_*$. And
2455: internal asymptotic solution has as the leader term the
2456: constant $U_*$. It leads to the fact that the fast oscillating
2457: solution degenerates at the point $U_*$ and it allows us to
2458: match constructed solutions in the different domains. This
2459: reason allows us to evaluate the value $E(t_*)=E_*$
2460: $$
2461: E_*={1\over2}|U_*|^4-t|U_*|^2-(U_* + U_*^*).
2462: $$
2463: By substituting these values we obtain
2464: $E_*={3\over4}\left({1\over2}\right)^{1/3}$.
2465: At the point $U_*$ all three roots of the equation
2466: $$
2467: 2y^3+(2E+t_*^2)y^2 + 2t_*y+1=0
2468: $$
2469: coalesce. It means that the integrand in the equation
2470: (\ref{u0s}) has the singularity of the order of $3/2$ and
2471: elliptic integral in (\ref{u0s}) degenerates. The obtained
2472: values of the constants $E_*$ and $t_*$ allow us to evaluate
2473: the adiabatic invariant $\s$. To obtain this value it is
2474: necessary to evaluate the integral
2475: \bb
2476: I_*=\int_{\G(t_*)}u^* du=\s_*. \label{sigma}
2477: \ee
2478: As result the equation (\ref{inv}) defines the function $E(t)$
2479: for our confluent solution. Thus to obtain the leader term of
2480: the asymptotics (\ref{as4}) completely it is necessary to
2481: determine the phase shift $S_0$ in the formula ({\ref{u0s}).
2482: It can be evaluated by matching of the asymptotics
2483: (\ref{as4}) and internal asymptotics from previous section.
2484:
2485:
2486: \subsection{The domain of suitability of confluent asymptotic
2487: solution as $t\to t_*-0$}
2488:
2489:
2490: To match the internal asymptotic expansions with the fast
2491: oscillating expansion it is necessary to know where the last
2492: one is valid as $t\to t_*$. From the formula (\ref{period})
2493: it is easy to see that the function $S'$ is defined by the
2494: value inverse to an integral
2495: $$
2496: K=\int_{\G(t)}{dy\over\sqrt{2y^3+(2E+t^2)y^2+2ty+1}}.
2497: $$
2498: So to obtain the domain of suitability of the fast oscillating
2499: solution it is necessary to determine the order of
2500: singularity of the leader term of the asymptotics of this
2501: integral as $t\to t_*-0$.
2502: \par
2503: The integral $K$ contains two parameters $t$ and $E$. The
2504: function $E=E(t)$ is defined by equation (\ref{inv}). In the
2505: neighborhood of the point $t_*$ we use the following notation
2506: $$
2507: t=t_*+\mu,\quad E=E_*+\d.
2508: $$
2509: Let us change the variable in the integral
2510: $$
2511: y=-\bigg({1\over2}\bigg)^{1/3}+\z.
2512: $$
2513: Then the integral $K$ looks like
2514: \ba
2515: K=\int_{\G(t)}d\z\bigg(2\z^3+
2516: \z^2\bigg[2\d+6\bigg({1\over2}\bigg)^{1/3}\mu+\mu^2\bigg]+
2517: \\
2518: \z\bigg[-2^{5/3}\d-4\mu-2^{2/3}\mu^2\bigg]+\\
2519: \bigg[2^{1/3}\d+3\bigg({1\over2}\bigg)^{1/3}\mu-2^{2/3}\mu+
2520: \bigg({1\over2}\bigg)^{2/3}\mu^2\bigg]\bigg)^{-1/2}. \ea
2521: \par
2522: The curve $\G(t)$ as $t=t_*$ passes through a point $\z=0$.
2523: At this point the integrand at the moment $t=t_*$ has
2524: nonintegrable singularity and integral diverges.
2525: \par
2526: The integrand and curve $\G(t)$ give us the information that
2527: to evaluate the main term of the asymptotics as $t\to t_*-0$
2528: of this integral it is enough to evaluate the integral only on
2529: the small segment of the curve $\G(t)$ near the point $\z=0$.
2530: \par
2531: To make this evaluations it is convenient to pass to the
2532: integral on a small segment of the real axis and to write the
2533: complex variable $\z$ in the form
2534: $$
2535: \z=\xi+i\eta,\quad \xi,\eta\in {\hbox{R}}.
2536: $$
2537: In a small neighborhood of point $\z=0$
2538: the imaginary part of the $\z$ can be expressed by real part
2539: by formula
2540: $$
2541: \eta_{\pm}(\xi)=
2542: $$
2543: $$
2544: \pm\sqrt{2^{1/3}+2^{2/3}\xi-\xi^2+\mu-
2545: 2^{1/3}\sqrt{1+3\bigg({1\over2}\bigg)^{1/3}\mu+2^{1/3}\d+
2546: 2^{-2/3}\mu^2+2^{4/3}\xi}}.
2547: $$
2548: Now if we assume that $\xi$, $\mu$ and $\d$ are small then
2549: we obtain \ban \eta_{\pm}^2=
2550: -2^{-2/3}\xi^3+2^{-1/3}\bigg[3\bigg({1\over2}\bigg)^{1/3}\mu+
2551: 2^{1/3}\d\bigg]\xi+\bigg[-\mu/2-2^{1/3}\d\bigg]+
2552: \nonumber\\
2553: O(\xi^4)+O(\xi^2\mu)+ O(\xi^2\d)+
2554: O(\xi\d)+O(\xi\mu)+O(\mu^2)+O(\d^2)+O(\mu\d). \label{eta-as}
2555: \ean
2556: The function $\d(\mu)$ stays undetermined in this
2557: formula.
2558: \par
2559: The dependence $E(t)$ and hence $\d(\mu)$ can be obtained from
2560: (\ref{inv}). Let us write this equation by integrals on
2561: segments
2562: \bb
2563: \int_{x_l}^{x_+}dx\sqrt{t-x^2+\sqrt{t^2+2E+4x}}-
2564: \int_{x_l}^{x_-}dx\sqrt{t-x^2-\sqrt{t^2+2E+4x}}=I_*. \label{E}
2565: \ee
2566: Where $x_l=-(t^2+2E)/4$, $x_+$ is the right real root of the
2567: equation
2568: \bb x^4-2tx^2-4x-2E=0, \label{p4} \ee
2569: and $x_-$ is the left one.
2570: \par
2571: To construct the asymptotics of the function $E(t)$ as $t\to
2572: t_*-0$ it's necessary to write the asymptotics of the roots
2573: of the equation (\ref{p4}). This equation in the investigated
2574: domain has two real roots and two complex conjugate roots. So
2575: we write
2576:
2577: $$
2578: (x-x_+)(x-x_-)((x-m)^2+n^2)=0.
2579: $$
2580: Where $x_-$, $x_+$, $m$ É $n$ are real functions with
2581: respect to $t$. They satisfy to Vieta equations
2582:
2583: \ba
2584: x_-x_+(m^2+n^2)=-2E,\\ -(x_-+x_+)(m^2+n^2)-2mx_-x_+=-4,\\
2585: m^2+n^2+x_-x_++2m(x_-+x_+)=-2t,\\x_-+x_++2m=0.
2586: \ea
2587: \par
2588: We seek a solution of these system of equations in the form
2589: $$
2590: x_-=U_*+\a,\quad x_+=-4U_*+\b,\quad m=U_*+m_1.
2591: $$
2592: The substituting of the values $U_*$, $E_*$, $t_*$ leads us to
2593: \ba
2594: 2^{2/3}[3m_1^2-n^2-\mu]-2m_1[-m_1^2+n^2+\mu]=0,\\
2595: 2^{1/3}[6m_1^2-2n^2-\mu]+2^{3/5}m_1[\mu-3m_1^2-2n^2]+
2596: \\
2597: +[m_1^2+n^2][-2\mu+3m_1^2-n^2]=-2\d. \ea
2598: \par
2599: The asymptotics of the solution for this system of the
2600: equation is seeking in the form \ba
2601: m_1=\m_1\sqrt{-\mu}+\m_2\mu+\dots,\\
2602: n=\n_1\sqrt{-\mu}+\n_2\mu+\dots,\\
2603: \d=\d_1\mu+\d_2(-\mu)^{3/2}+\dots. \ea By substituting into
2604: the system we get
2605: $$
2606: 3\m_1^2-\n_1^2=1,\quad \d_1=-\bigg({1\over2}\bigg)^{2/3}.
2607: $$
2608: Thus in the first corrections we did not determine a relations
2609: between $\m_1$ and $\n_1$. This relation can be determine from integral
2610: equation (\ref{E}).
2611: \par
2612: Let us construct the main term of the asymptotics of the
2613: integral $K$. In the small neighborhood of the point $\z=0$
2614: the curve $\G(t)$ intersects of the real axis. So the integral
2615: over a small neighborhood of the point $\z=0$ can be
2616: represented in the form
2617: \ba K_0\sim\int_\a^{\xi_{max}}{\pt_\xi\z_+(\xi)
2618: d\xi\over\sqrt{2\z_+^3-2[-2^{4/3}+2]\mu\z_+
2619: +\a_+(-\mu)^{3/2}}}+\\
2620: \int_\a^{\xi_{max}}{\pt_\xi\z_-(\xi)
2621: d\xi\over\sqrt{2\z_-^3-2[-2^{4/3}+2]\mu\z_-
2622: +\a_-(-\mu)^{3/2}}}, \ea where $\xi_{max}$ is the null of the
2623: radicand, $\z_{\pm}=\xi+i\eta_{\pm}(\xi)$, $\a_{\pm}$ are
2624: constants. The parameter $\a$ is chosen such that
2625: $\a(-\mu)^{-1/2}\gg1$ and $\a\ll1$ as $\mu\to-0.$
2626: \par
2627: The asymptotics of the function $\eta_{\pm}$ and the explicit
2628: formula for the main term of the integral $I_0$ and change of
2629: the variable $\xi=(-\mu)^{1/2}\nu$ in integral
2630: lead us to the formula
2631: $$
2632: K_0=O((-\mu)^{-1/4}).
2633: $$
2634: Therefore
2635: $$
2636: S'=O((-\mu)^{1/4}),\quad \hbox{as}\ \mu \to -0.
2637: $$
2638: By the similar evaluation one can show that
2639: $$
2640: \pt_E I=O((-\mu)^{-1/4}),\quad \pt_E S=O((-\mu)^{1/4}),
2641: {\hbox{as}}\,\,\mu \to -0.
2642: $$
2643: Therefore $\phi'=O((-\mu)^{1/2})$.
2644: \par
2645: By this way we obtain the formula
2646: $$
2647: \pt_t\stackrel{0}{U} = O(\mu^{-1})\quad
2648: {\hbox{as}}\,\,\mu\to-0.
2649: $$
2650: \par
2651: Using $u_1$ and $u_2$ --- solutions of the homogeneous
2652: equation one can obtain that
2653: $$
2654: \stackrel{1}{U}=O((-\mu)^{-3/2}).
2655: $$
2656: \par
2657: As a result we obtain that the fast oscillating asymptotics
2658: is valid as $(t_*-t)\ve^{-2/3} \gg 1$. It is easy to see that
2659: the domains, where the constructed asymptotic solutions are
2660: valid, are intersected.
2661: \par
2662:
2663: \subsection{Matching of the asymptotics}
2664:
2665: \par
2666: Let us match the fast oscillating asymptotics and internal
2667: asympotics
2668: (\ref{infs}), (\ref{weia}) into the transition layer.
2669: \par
2670: The main term of the fast oscillating asymptotics is defined as
2671: implicit function by the formula
2672: $$
2673: t_1=-iS'\int^{U_*+W}_{y\in\G(t)}{dy\over -|y|^2y-ty+1}.
2674: $$
2675: \par
2676: This integral can be represented in the form as
2677: $t\to t_*+0$
2678: $$
2679: t_1=-iS'\bigg(\int_{\G(t_*)}^{U_*+W}{dy\over
2680: -|y|^2y-t_*y+1+O(t-t_*)y} \,+\,
2681: $$
2682: $$
2683: \int_{\d\S_{\G}}^{U_*+W}
2684: dy\wedge dy^*\pt_{y^*}\bigg[{1\over
2685: -|y|^2y-t_*+1+O(t-t_*)y}\bigg]\bigg).
2686: $$
2687: Where $\d{\S_{\G}}$ is a figure bounded by the curves $\G(t)$
2688: and $\G(t_*)$.
2689: \par
2690: The function $U_*+W$ is periodic. In the case the following
2691: inequalities are
2692: $$
2693: {\d\s_{\G}\over W^2}\ll1\quad \hbox{(where}\,\, \d\s_{\G} \hbox{--
2694: square of the figure }\,\,\d\S_{\G}) \quad {t-t_*\over W^{7/2}}\ll1,
2695: $$
2696: we obtain the following representation in some neighborhoods
2697: of the point $t_{(k)}$
2698: $$
2699: \stackrel{0}{U}\sim U_*+W(t_{(k)}),\quad
2700: $$
2701: Where
2702: $$
2703: W(t_{(k)})={-2\over (t_{(k)}-iU_*)^2}.
2704: $$
2705: The argument of the function $W$ can be represented in the form
2706: $$
2707: {t_1\over S'}\sim
2708: {S(t_{(k)})+S'(t_{(k)})(t-t_{(k)})+o(t-t_{(k)})\over
2709: \ve(S'(t_{(k)})+O(S''(t_{(k)})(t-t_{(k)}))} \,+\,
2710: {\phi(t_{(k)})\over S'(t_{(k)})},\quad {\hbox{as}}\,\,\,\,
2711: t\to t_{(k)}.
2712: $$
2713: By matching with the main term of the internal asymptotics we
2714: obtain
2715: $$
2716: {S(t_{(k)})\over \ve S'(t_{(k)})} \,+\, {t-t_{(k)}\over \ve}
2717: \,+\, {\phi(t_{(k)})\over S'(t_{(k)})}\sim \th_k.
2718: $$
2719: \par
2720: This matching with the second internal asymptotics gives us
2721: that this fast oscillating solution also can be matched with
2722: the intermediate expansion. The constants $\phi_0$ and
2723: $\phi_1$ evaluate by this way. These constants define the
2724: phase shift of the fast oscillating solution. But this kind of
2725: evaluations are not made in this work.
2726: \par
2727: We are grateful to L.A. Kalyakin, B.I. Suleimanov for
2728: stimulating discussions. And we also thank N. Enikeev for
2729: his helpful and comments.
2730:
2731:
2732:
2733:
2734: \begin{thebibliography}{cc}
2735: \bibitem{K-B}
2736: Krylov N.M., Bogolyubov N.N.. Introduction into nonlinear
2737: mechanics. Izdatel'stvo AN USSR, 1937.
2738: \bibitem{Andr}
2739: Andronov A A and Chaikin C E, Theory of Oscillations,
2740: Princenton University Press, Princenton, NJ, 1949.
2741: \bibitem{Arn}
2742: Arnold V I, Afraimovich V S, Il'yashenko Yu S and Shil'nikov
2743: L P, Theory of Bifurcations, in Results of the Science and
2744: Engineering. Modern Problems of Mathematics. Fundamental
2745: Directions, Nauka, Moscow, 1986, V.5, 9--218.
2746: \bibitem{Z-S}
2747: Zasslavskii G.M., Sagdeev R.Z.. Introduction into nonlinear
2748: physics. Moscow, Nauka, 1988.
2749: \bibitem{Hab1}
2750: Haberman R.. Slowly varying jump and transmision phenomena
2751: associated with aljebraic bifurcation problems. SIAM J. Appl.
2752: Math., v.37, n1, pp. 69-106.
2753: \bibitem{ok1}
2754: O.M.Kiselev. Asymptotic approach for the rigid condition of
2755: appearance of the oscillations in the solution of the
2756: Painlev\'e-2 equation.
2757: http://xxx.lanl.gov/solv-int/9902007
2758: \bibitem{Dim-Hab}
2759: Diminnie D.C., Haberman R.. Slow passage through a
2760: saddle-center bifurcation. J. Nonlinear Sci., v.10,
2761: pp.197-221(2000).
2762: \bibitem{ok2}
2763: Kiselev O.M., Hard loss of stability in Painlev\'e-2
2764: equation. J. of Nonlin. Math. Phys., 2001, v.8, n1, pp.65-95.
2765: \bibitem{Timof}
2766: Timofeev A.V.. To a question about constancy of an adiabatic
2767: invariant as changing of the type of the motion. ZhETP, 1978,
2768: v.75, n4, pp.1303-1308.
2769: \bibitem{Neisht}
2770: Neishtadt A.I., About change of adiabatic invariant as passing
2771: through a seperatrix. Fizika plazmy, 1986, v. 12, n8, pp.
2772: 992-1001.
2773: \bibitem{Bour-Hab1}
2774: F.J. Bourland, R. Haberman. Separatrix crossing:
2775: time-invariant potentialswith dissipation. SIAM, J.Appl.
2776: Math., v.50,n6,pp.1716-1744.
2777: \bibitem{Hab-Ho}
2778: R. Haberman, E.K.Ho. Boundary of the basin of attraction for
2779: weakly damped primary resonance. J. Appl. Mech., v.62,
2780: pp.941-946.
2781: \bibitem{Hab}
2782: Haberman R.. Nonlinear transition layers-- the second Painleve
2783: transcendent. Stud. Appl. Math., v.57, pp.247-270.
2784: \bibitem{Kuz}
2785: Kuzmak G E, Asymptotic Solutions of Nonlinear Differential
2786: Second-Order Equations with Variable Coefficients, {\it Appl.
2787: Math. and Mech.}, 1959, V.23, N~3, 515--526.
2788: \bibitem{Bour-Hab2}
2789: Bourland F.J., Haberman R., The modulated phase shit or strongly
2790: nonlinear, slowly varying and weakly damped oscillators. SIAM J.
2791: Appl. Math., 1988, v.48, n3, pp.737-748.
2792: \bibitem{I}
2793: Il'in A M, Matching of Asymptotic Expansions of Solutions of
2794: Boundary Value Problem, AMS, 1992.
2795: \bibitem{H-S} P. Holmes, D.
2796: Spence. On a Painleve-type boundary-value problem. The
2797: Quaterly Journ. of Mech. and Appl. Math., 1984, v.37, n4,
2798: p.525-538.
2799: \bibitem{G-L}
2800: Gromak V I and Lukashevich N A, Analytical Properties of
2801: Solutions of the Painlev\'e Equations, Minsk, 1990.
2802: \bibitem{Kit}
2803: Kitaev A V, Elliptic Asymptotics First and Second
2804: Painlev\'e-Transcendents, Usp. Matem. Nauk, 1994, v.49, n1,
2805: 77--140.
2806: \bibitem{Abramowitz}
2807: Handbook of mathematical functions. Ed. by M. Abramowitz and
2808: I. A. Stegun, 1964.
2809:
2810:
2811: \end{thebibliography}
2812:
2813: \end{document}
2814: